AbstractWith the development of deep underground space, there is an ever-growing demand for the excavation of deep foundations in the urban zone. The diaphragm wall is the most commonly used retaining and waterproof structure. Therefore, the quality of the diaphragm wall plays an essential role in the environmental impact and security risk control of deep foundation excavation. Consequently, it is required to develop practical ways to detect defects in diaphragm walls. It is critical to localize defects before excavation and characterize them in detail for proper countermeasures. Unfortunately, up to now, the traditional detection methods are not yet available for professional use. This paper presents a contribution to filling this gap based on cross-hole sonic logging (CSL) and ground-penetrating radar (GPR). First, CSL is carried out prior to excavation through detection tubes preinstalled in the wall. Second, Bayesian inversion is used to obtain a high-quality two-dimensional (2D) visualization of the defects, and the depth and width can be obtained before excavation. Third, GPR is applied to precise detection when excavating to the depth of the defect. Finally, by employing an advanced bifrequency back-projection method, the correlation between two domain frequency radar data is used to obtain the three-dimensional (3D) characteristics of the defects. The proposed method was used at a 40-m level of deep excavation in Shanghai, China, and allowed engineers to determine the location and 3D characteristics of defects. An identified analysis of the defects with this method showed a reasonably good match with manual observation after excavation.