Abstract: Cosmic-ray muon imaging is a new non-destructive technique that uses highly penetrating natural muon to reconstruct objects with high atomic number (Z) or subjects hidden in something by applying muon scattering or transmission technique, respectively. Up till now, muography mainly focused on objects with high Z or large volume. But few studies have been reported on objects at small scale with low or medium Z in China. A method that aims to reconstruct 3D image for objects at small scale as well as low or medium Z has been brought out in this paper, which uses the amount of muonic secondary particles and coinciding muon trajectories in different directions to reconstruct objects. A simulation about how muonic secondary particles generate and their performances based on Geant4 toolkit was carried out. The results of this simulation show that the secondary particles are mainly composed of secondary electrons and gamma photons. And as the thickness of the objects increases along one certain direction, the amount of secondary particles increases approximately linearly. Therefore, the amount of them carries the shape and location information of objects. Furthermore, the energy of muonic secondary particles is mainly beyond 0.1 MeV, so secondary particles can be distinguished with the background noise in the experimental environment. After selecting coincide muon trajectory through injected cosmicray muon and its secondary particles coincide detection technique, the 3D images of three 3 cm×3 cm×3 cm cubes made up from different materials (iron, water, and lead) were reconstructed based on limitedangle computed tomography algorithm. It can be seen from the imaging results that the reconstructed image of iron has best quality. Both water and lead have some voids in their reconstructed images, but the voids of water is apparently less than those of lead. Thus, this imaging method is fit for imaging small scale objects with low or medium Z and the imaging effect of objects at the small scale with medium Z is especially good. A reconstruction of complex objects which is defined as “USC” letter model has also been made. The reconstructed image can match with the size and location of the geometry model in Geant4 and can also distinguish the intervals of 0.8 cm in the letter “S”. A much better imaging accuracy will be achieved by optimizing the mathematical model of the algorithm. This method extends the application area of muon tomography.