To　tackle　the　challenge　of　multipath　effects　in　FMCW　radar　systems　used　for　indoor　tracking,　which　can　lead　to　ghost　images　and　impede　target　detection,　this　paper　presents　a　two-stage　approach　to　address　this　issue.　First,　a　linear　identification　method　based　on　the　Random　Sample　Consensus　(RANSAC)　algorithm　is　applied　in　the　Range-Doppler　domain　to　differentiate　between　the　true　target　and　second-order　ghost　images.　By　leveraging　the　inherent　mirror　symmetry　relationship　between　the　true　target　and　ghost　images,　a　least　squares　method　is　then　proposed　to　optimally　estimate　the　wall　position.　Experimental　results　demonstrate　the　efficacy　of　this　approach,　achieving　a　wall　position　estimation　accuracy　with　an　overall　error　of　1.1　degrees,　thereby　enabling　reliable　wall　location　estimation.