It is known that the non-contact measurements of the flexible mechanism vibration by the machine vision method are susceptible to the environmental interference, therefore, this paper proposes a non-contact measurement method, the maximal between-cluster variance algorithm based on the image binarization. Firstly, due to the drawbacks of the traditional twodimensional maximal between-cluster variance algorithm, such as the misclassification and the complicated computation, the twodimensional histogram partitioning method and the variable step size iteration method are used to improve the traditional twodimensional maximal between-cluster variance algorithm, then the Lena image and the flexible cantilever beam are taken as examples to carry out the simulation analysis and the vibration displacement measurement experiments, respectively. The results show that the non-contact measurement method based on the improved maximal between-cluster variance algorithm significantly improves the image segmentation effect, and the segmentation time is 10% of the traditional maximal between-cluster variance algorithm. With the vibration displacement of the end of the flexible cantilever beam obtained by the simulation as the criterion for the measuring accuracy, the measurement result of the improved two-dimensional maximal between-cluster variance algorithm is found superior to the traditional two-dimensional maximal between-cluster variance method and the piezoelectric method, with feasibility and reliability.

vibration of flexible structure  /  non-contact measurement  /  image binarization  /  maximal between-cluster variance  /  two-dimensional histogram  /  variable step size iteration