Using lightweight convolutional neural network to track vibrationdisplacement in rotating body video

The non-contact visual displacement measurement methods can detect problems such as rigidbody deformation and structural wear of structural bodies due to long-term service life.However, the traditional visual vibration measurement methods are limited by the inherentsampling constraint of ordinary image sensors, which cannot achieve efficient vibration targetidentification and correlation tracking for rotating body targets in low-resolution videos underthe premise of high sampling rate. Therefore, this paper uses high-speed industrial camera asthe acquisition medium, introduces deep convolutional neural networks into the field of visualvibration measurement, on this basis, weighs the target detection accuracy and displacementtracking speed, uses the designed lightweight convolutional neural networks model to vibrationdisplacement tracking in rotating body videos. We ensure that the computational efficiency andparameter quantity of the convolutional networks are solved with low loss accuracy. First ofall, we take the lightweight convolutional neural network as the backbone network, replace thestandard convolutional neural network with depthwise separable convolution and pointwiseconvolution. Considering the speed and accuracy advantages of deep learning algorithms forvideo object tracking, we estimate the heat map, object center offsets and bounding box sizesuse an anchor-free detection algorithm on the network framework. In order to prevent theloss of vibrating target identity in rotating body videos, we use re-identification(re-ID) methodto strengthen the correlation of target displacement between adjacent frames. In experimentswith traditional visual vibration measurement and recent deep learning measurement methodsfor performance testing, the network model we designed can demonstrate absolute advantagesthat traditional convolutional neural networks do not have. On the one hand, our comparison oftime-frequency characteristics at different speeds shows that the vibration displacement curveregressed by the lightweight convolutional neural network has a high degree of fit with thedisplacement signal obtained by the eddy current sensor;on the other hand, when the targetobject is blurred, the generalization ability of our algorithm is proved, which also reflectsthe engineering application value of visual vibration measurement in the field of vibrationdisplacement tracking of rotating bodies.