To address the challenge of high-precision inversion for 3D subsurface magnetic structure distribution under complex terrain and geological body conditions, an unstructured grid magnetization inversion method constrained by magnetic anomaly ΔT correlation imaging was proposed. Tetrahedral unstructured grids were adopted to enhance the fitting accuracy of complex topography and geology. Correlation imaging was utilized to identify key regions, thereby enabling model dimensionality reduction, significantly compressing the solution space scale, and improving inversion efficiency. For the key regions after dimensionality reduction, depth-weighted constraints and physical property range constraints are introduced for regularized inversion, which can effectively focus on the target regions and improve the inversion accuracy. The preconditioned conjugate gradient method was adopted to solve the objective function, and the OpenMP parallel algorithm was introduced to further boost the inversion efficiency. The advantages of this method in improving the accuracy and efficiency of magnetization inversion were verified by model tests, and when it is applied to the measured aeromagnetic anomaly data in the McFaulds Lake area, Ontario, Canada, the correctness and effectiveness of the method are further confirmed. This method provides a new technical means for the magnetization inversion of complex geological bodies and possesses significant innovative value and application potential.