In this paper, we propose a new tomographic reconstruction method,called IOD-PVRMPP, to reconstruct volume of 3D particles fromimages of 2D particles provided by the Tomographic Particle ImageVelocimetry (Tomo-PIV) technique. Our method, based on objectprocesses, allows to solve the problem in a parsimonious way. Itfacilitates the introduction of prior knowledge and solves memoryproblem which is inherent to voxel based approaches used by classi-cal tomographic reconstruction methods. The reconstruction of a setof 3D particles is obtained by minimizing an energy function whichdefines the object process. To this aim, we use a simulated anneal-ing algorithm based on Reversible Jump Markov Chain Monte Carlo(RJMCMC) method. To speed up the convergence of the simulatedannealing, we develop an initialization method which provides theinitial distribution of 3D particles. To do that, we proceed by de-tecting 2D particles located in projection images. Using syntheticdata, we show that IOD-PVRMPP method gives better results thanMinLOS-MART method in different seeding densities.