Non-line-of-sight imaging technology is a new technique to detect hidden objects, which cannot be observed directly. The 3D reconstruction algorithm in non-line-of-sight imaging is mainly the back projection method. Because the back projection algorithm needs to cover the data in the Angle range of 180° to obtain a good reconstruction result, while non-line-of-sight imaging technology can only obtain partial (sparse) angle data, the back projection algorithm cannot obtain an ideal reconstruction result. To solve the sparse data problem, the system adopts the APD array to acquire data quickly, and the simultaneous algebraic reconstruction technique (SART) is proposed to reconstruct 3D image of non-line-of-sight objects. Experimental results show that the proposed approach can remove the artifacts and noise better than in the back projection algorithm on spare angle data. By calculating the peak signal-to-noise ratio (PSNR) and structural similarity SSIM, in the second and third iterations, 1.3928, 2.4663 and 0.1198, 0.2312 are improved respectively, it is shown that this method can effectively improve the image reconstruction accuracy of non-line-of-sight objects.