Monitoring defects such as voids and foreign inclusions in building walls is critical for ensuring structural safety. Conventional methods like infrared thermography suffer from environmental interference, impact-echo techniques lack precise localization, and ultrasonic testing offers limited visualization. To address these limitations, we design a wall defect detection system based on Compton backscattering imaging and conduct simulation studies using Geant4. The detection system employs a radiation source emitting a fan-beam collimated by a front collimator to scan the wall model. Backscattered photons generated via Compton scattering are then collimated by a rear collimator and detected by a 256 × 256 pixel NaI(Tl) scintillator array detector. The system reconstructs cross-sectional density distribution images from scattered photon counts. After acquiring multiple layers of scans, image correction, noise reduction (Gaussian and median filtering), and threshold segmentation are applied, followed by three-dimensional (3D) reconstruction using volume rendering to extract structural details of internal defects. Geant4 simulations demonstrate that the system accurately identifies defect materials in concrete walls, with a positional error of less than 2%, a 3D reconstruction resolution of 1.0 mm, and both volume and surface area reconstruction errors below 4%.