Laser Doppler Velocimetry(LDV) is able to measure the velocity of vehicle accurately for high precision navigation in the area of aerospace. The key of the measurement is to detect the Doppler signal successfully, so weak Doppler signal detection is one of the most important technologies due to the presence of noise. An adaptive signal detection algorithm in frequency domain was proposed. Noise was leached from useful signal by band stop filter and the adaptive threshold could be obtained according to the statistical properties of the left noise. Comparisons of simulation and experiments with the fixed threshold algorithm were taken, and the results show that the frequency domain adaptive threshold detection is an effective method to improve the performance of LDV. Under the condition of keeping the low Constant False Alarm Rate (CFAR), the Doppler signal can be completely detected when the Signal-to-Noise Ratio(SNR) is higher than -9 dB, and it is insensitive to the intense of noise, simple and applicable in many area.