Concrete is widely utilized in engineering due to its stability, durability, and moldability. Among various forms, layered concrete structures are extensively employed in critical infrastructure owing to their clear functional zoning and construction convenience. However, such structures are often susceptible to delamination defects during service under external loads and complex environmental conditions, significantly compromising structural safety and durability. To address the issues of inaccurate localization and poor imaging quality in existing ultrasonic imaging of delamination defects, this paper proposes an ultrasonic imaging method based on statistical-feature root-mean-square (SRMS). This method integrates an adaptive statistical weighting mechanism, optimizing the focusing quality of ultrasonic images and the representation of defect echoes from the dual dimensions of propagation time-delay and channel weighting. It not only employs a root-mean-square (RMS) velocity model to calculate the equivalent acoustic velocity to improve localization accuracy, but also quantifies the coherent stability of local signals, thereby effectively suppressing lateral and axial energy divergence in the target region. Simulations and experimental validations on delamination defects in ballastless tracks demonstrate that the SRMS method achieves detection errors as low as 6.0 mm and 0.5 mm in lateral dimension and longitudinal localization, respectively. Compared with the traditional delay-and-sum (DAS) and RMS methods, the lateral detection errors are reduced by 129.5 mm and 93.5 mm, while the longitudinal localization errors are reduced by 2.0 mm and 7.5 mm. Furthermore, this method effectively compresses the longitudinal main-lobe width to 37.5 mm, achieving reductions of 17.5 mm and 16.5 mm compared to the DAS and RMS methods, respectively. In terms of imaging quality, because defect echoes are significantly enhanced and background noise is effectively attenuated during the superposition process, the contrast ratio (CR), contrast-to-noise ratio (CNR), and peak signal-to-noise ratio (pSNR) of the SRMS method reach 17.75 dB, 15.95 dB, and 56.23 dB, respectively. These metrics represent enhancements of 29%, 18%, and 164% over the DAS method, and 25%, 15%, and 173% over the RMS method. The proposed SRMS method not only improves the detection accuracy of delamination defects in layered concrete structures but also enhances imaging quality in complex environments.