Physics‐Embedded Full‐View Scattering Reconstruction With Sparse Local Observations

ABSTRACT

The three‐dimensional (3D) electromagnetic scattering field acts as a distinctive physical fingerprint encapsulating the intrinsic morphological and material properties of objects, which is essential for precise target identification and characterization. However, real‐time acquisition of such full‐view information is fundamentally constrained by the trade‐off between sampling density and measurement efficiency. Here, we introduce a physics‐embedded compressive sensing framework capable of reconstructing complete scattering fields only from sparse measurements. This framework utilizes a metasurface‐based generalizable physical proxy, from which the acquired data encapsulates the statistical priors of scattering behavior. Leveraging this data, we generate a versatile overcomplete dictionary of electromagnetic wave modes, forming a sparse basis for the compressive sensing algorithm to rigorously extrapolate global field distributions from limited local measurements. Both simulations and experiments demonstrate robust field recovery of complex targets under sparse sampling with an average cosine similarity above 0.87. By effectively bridging local sparse sensing with global field inference, our work establishes a generalizable, physics‐aware sensing paradigm, with promising implications for holographic radar, non‐destructive evaluation, and intelligent robotic perception.