Abstract: To improve the target detection accuracy of millimeter wave radar, a radar multi-target detection algorithm that combines three-dimensional fast Fourier transform (3D-FFT) with improved orthogonal matching pursuit (IOMP) is proposed in this paper. This work addresses the issues of low detection speed and poor angle parameter estimation accuracy encountered by traditional 3D-FFT algorithms under conditions of low chirp counts and limited receive antenna elements. The preliminary parameter estimation and spatial filtering are performed initially using the 3D-FFT, then the IOMP method is employed to conduct sparse reconstruction within the regions of interesting, which achieves higher precision in velocity and angle estimation. Verifications through both simulation and experimental data demonstrate that the proposed algorithm yields a one to two orders of magnitude improvement in velocity and angle estimation accuracy compared to other methods. Meanwhile, the overall computational complexity is significantly reduced, enhancing real-time processing capabilities. With superior detection performance and reduced computational overhead, the proposed algorithm exhibits strong potential for engineering applications, and lays a solid foundation for the development of subsequent millimeter wave radar multi-target tracking systems.