Laser heterodyne detection (LHD) is a key velocimetry technique that provides better accuracy and sensitivity than direct laser detection. However, random phase noise can be introduced by the surface topography of the moving target undulation or atmospheric turbulence during transmission. The random phase noise causes the target echo to undergo decoherence, resulting in degradation of the signal-to-noise ratio (SNR). Here, we propose a novel LHD method based on photon time–domain differential detection. It can infer the heterodyne spectrum of the target echo and the local oscillator light from the time intervals of the photon arrival. The time interval statistic is a relative quantity, which can effectively avoid the effect of random phase noise in LHD. With our method, the SNR of LHD can be improved in application scenarios where the target echo is decoherent. We developed a complete solution model for acquiring the heterodyne spectrum based on photon time–domain differential detection and performed proof-of-principle experiments. The experimental results show that in the presence of random phase noise, the SNR and velocity measurement error of our method are significantly better than that of the conventional method, and the larger the phase noise is, the more the SNR and velocity measurement error of our method are improved. Moreover, along with the increase in phase noise, the SNR of our method is basically unchanged, which also indicates that our method is not affected by random phase noise. This advantage is significant for photon-level weak echoes that require long detection times to be detected. [ABSTRACT FROM AUTHOR]