This study aims to detect REE geochemical anomalies in relationship to Iron-apatite ores utilizing an image Fusion based on Deep Learning (FDL). The geochemical data was modeled for elements related to Iron-apatite mineralization using multi b Spline B. The results were fusioned in applying the Deep learning method based on pre-trained networks. Wavelet-Number (WN) fractal model classified the best results based on the combination of a two-dimensional Discrete Wavelet Transformation (DWT) signal analysis and a Concentration-Area (C-A) fractal modeling. Sym8 carried the DWT as a selected wavelet pattern for REE based on Stream sediment samples collected from the Tarom region (NW Iran). In addition, the DWT was decomposed by wavelet coefficients at five levels. Furthermore, the DWT data were classified using a fractal-wavelet model to delineate REE anomalies from background levels in this region. Overlayed with the catchment basins model and weighted using the upstream and downstream parts. As a result, the prominent REE source anomalies are located in the southern parts of the study area. The results obtained by the proposed fractal-wavelet modeling are in connection with field check anomaly samples and the rock samples collected from the Iron-Apatite ore deposits.