This article presents a machine-perception method based on eddy current density (ECD) field induced in a plate for reconstructing nonuniform electrical conductivity. Formulated as a two-stage inverse problem using distributed current source models, the ECD field is reconstructed from its magnetic flux density and followed by solving an inverse ECD model to estimate the conductivity fields of the plate and its features. Along with a parametric study, the effects of different sensor and coil configurations on the reconstruction accuracy are numerically investigated. The proposed method is evaluated experimentally by comparing two methods: mechanical scanning using a single-sensor and a linear method with a sensitivity matrix. The findings demonstrate that using a sensor-array with a multicoil configuration effectively reduces the number of mechanical scans when making measurements, and eliminates variations below the coil center where the ECD is small.