The motivation for this work is the reconstruction of inclusions in complex media. The fine structure of the medium of propagation is here not available and is thus modeled as random. When the fluctuations are too strong, this precludes the use of standard imaging techniques that are based on the complete knowledge of the medium. We propose an alternative based on deterministic transport models that describe the propagation of the average of weakly random quadratic quantities in the wavefield. The crucial part of the approach is to quantify the error made by approximating weakly random processes by their average. Media with short-range and long-range correlations are considered and numerical simulations validating the theory will be presented.