The scattering centre model is well known as an approximation to provide a sparse abstraction of complex targets in radar applications. On the basis of the block-sparsity and high-frequency approximations, an improved scattering centre representation scheme is proposed for different angular observations of the land-based vehicles. For good insights into target physics and feature stability, the land-based vehicle is observed by airborne radar system in separate observation sections according to independent aspect angle regions. On the basis of that, the image patch of the land-based vehicle can be reconstructed with the measurements in each observation section. To reduce redundant information, compressive sensing followed by the CLEAN algorithm is exploited to extract the scattering centres from each image patch separately. Thus, the predominant scattering centres of the land-based vehicle can be grouped into several sub-structures with respect to the image patches. Numerical simulations are used to verify the validity and efficiency of the block-sparse representation (SR) scheme. The block-sparse representation scheme can be implemented with low computational complexity and flexibly extended to further research with multiple targets representation and applications in the radar field.