Instantaneous Rotation Speed Estimation Through Low-Cost Digital Imaging and Time-Frequency Analysis

Instantaneous rotation speed (IRS) estimation is desirable in a wide range of industrial fields for health assessment and system control of mechanical machinery. In this article, a low-cost visual measurement system is proposed to determine the IRS of a rotor through image processing and time-frequency analysis techniques. First, an image similarity signal is obtained by comparing sequential images with the reference image based on the zero-normalized cross-correlation algorithm. Then, a novel time-frequency analysis method, which deploys the Chirp-Z transform to refine the low spectral resolution of the short-time Fourier transform, is developed to analyze the image similarity signal and accurately determine the IRS. To overcome the difficulty of selecting the fundamental harmonic in the time-frequency map due to the multicomponent characteristic of the image similarity signal, the accelerated-KAZE feature detection and matching algorithm is employed. The measurement data collected on an experimental rig illustrate that the IRS can be estimated with a maximum error of ±0.5%. In addition, the measurement results confirm the superiority of this technique in dealing with images with motion blur and poor illumination conditions.