An improved measurement pixel selection method of GB-MIMO radar image applied for bridge vibration measurement

Ground-based multi-input multi-output (GB-MIMO) radar has significant advantages in bridge structural vibration parameter measurement due to its non-contact, high frame rate, and high-precision measurement characteristics. Due to the complexity of the environment, traditional methods of selecting pixels based on permanent scatterers (PS) cannot accurately obtain bridge structure pixels. This article proposes an improved classification method of the PS deformation curves to accurately select vibration pixels of bridge structures. Firstly, prescreen those PSs with small deformation amplitude as the first category. Secondly, for the screened PSs, the principal component analysis method is used to reduce the data dimension of their deformation curves to reduce the time consumption of clustering. Finally, the K-medoids algorithm is used to classify the screened PSs based on the dimensionality reduction data, and the Davis-Bouldin index is adopted to overcome the problem of local optimization. The experimental results of a large-scale bridge prove that the method proposed in this paper can effectively improve the speed of data processing while ensuring accurate pixel selection of the bridge structure.