Suspensions of semi-transparent particles such as polystyrene microparticles are commonly used as model systems in the study of micro-rheology, biology, and microfluidics. Holography is a valuable tool that allows one to obtain 3-D information for particle position and orientation, but forward reconstruction techniques often struggle to infer this information accurately for semi-transparent spheroids with an O(1) aspect ratio, since the lens effect from the particle introduces complex patterns. We propose a reconstruction method that uses image moment information to generate a mask over the sharp patterns from the lens effect and gives reasonable estimation of the 3-D position and orientation of the particle. The method proposed in this work uses the average particle geometry information to determine the process parameters and identify the appropriate detection zone. The average detection error for zc is less than 25% of the average particle thickness, and the average errors in the in-plane and out-of-plane orientations ϕ and θ are 2° and 4°, respectively. Our method provides comparable accuracy in the detection of the particle center of mass (xc, yc, zc) and in-plane orientation ϕ as a recent forward reconstruction method for semi-transparent particles proposed by Byeon et al. (H. Byeon, T. Go and S. J. Lee, Appl. Opt., 2016, 54, 2106-2112; H. Byeon, T. Go and S. J. Lee, Opt. Express, 2016, 24, 598-610). This method provides a clearly defined framework for identifying the particle's out-of-plane tilt angle θ. We finally demonstrate the applicability of the method to opaque, slender (aspect ratio AR ≫ 1) particles by analyzing the 3-D motion of E. coli cells from holographic video footage.