Convenient　and　helpful　defect　information　within　the　magnetic　field　signals　of　an　energy　pipeline　is　often　disrupted　by　external　random　noises　due　to　its　weak　nature.　Non-destructive　testing　methods　must　be　developed　to　accurately　and　robustly　denoise　the　multi-dimensional　magnetic　field　data　of　a　buried　pipeline.　Complete　ensemble　empirical　mode　decomposition　with　adaptive　noise　(CEEMDAN)　is　an　innovative　technique　for　decomposing　signals,　showcasing　excellent　noise　reduction　capabilities.　The　efficacy　of　its　filtration　process　depends　on　two　variables,　namely　the　level　of　additional　noise　and　the　number　of　ensemble　trials.　To　address　this　issue,　this　paper　introduces　an　adaptive　geomagnetic　signal　filtering　approach　by　leveraging　the　capabilities　of　both　CEEMDAN　and　the　salp　swarm　algorithm　(SSA).　CEEMDAN　generates　a　sequence　of　intrinsic　mode　functions　(IMFs)　from　the　measured　geomagnetic　signal　based　on　its　initial　parameters.　The　Hurst　exponent　is　then　applied　to　distinguish　signal　IMFs　and　reproduce　the　primary　filtered　signal.　SSA　fitness,　representing　its　peak　value　(excluding　the　zero　point)　in　the　normalized　autocorrelation　function,　is　utilized.　Ultimately,　optimal　parameters　that　maximize　fitness　are　determined,　leading　to　the　acquisition　of　their　corresponding　filtered　signal.　Comparative　tests　conducted　on　multiple　simulated　signal　variants,　incorporating　varied　levels　of　background　noise,　indicate　that　the　efficacy　of　the　proposed　technique　surpasses　both　EMD　denoising　strategies　and　conventional　CEEMDAN　approaches　in　terms　of　signal-to-noise　ratio　(SNR)　and　root　mean　square　error　(RMSE)　assessments.　Field　testing　on　the　buried　energy　pipeline　is　performed　to　showcase　the　proposed　method＇s　ability　to　filter　geomagnetic　signals,　evaluated　using　the　detrended　fluctuation　analysis　(DFA).