Abstract: With the increasing application of nuclear power and nuclear technology in energy, military, industrial and medical fields, a large number of different types of radioactive solid waste have been producing in nuclear fuel plant, reactor, military facilities, hospitals and research institutions. The rapid and accurate determination of the distribution, location and dosage of radioactive wastes is of great significance for safe operation of nuclear facilities, assessment and treatment. In view of the fact that fewer gamma camera projection data could be obtained in actual radiation source measurements, a sparse reconstruction technique based on compressed sensing was used to reconstruct the 3D distribution of the unknown sources inside the metal barrel. According to the reconstruction of two sets of projection data from the real sources and simulated sources experiments, the results show that the algorithm can accurately reconstruct the radiation sources' 3D position, shape, relative intensity in the sparse projection conditions, and achieves quantitative 3D reconstruction of the radiation source in the metal barrel.