Abstract: In order to improve the real-time detection capability of an electronic reconnaissance system for linear frequency modulation continuous wave(LFMCW)signals in non-stationary noise environment, a low complexity algorithm for segmental detection of signals was proposed. Setting the width of the window function based on the maximum chirp rate assumption, and dividing the intercepted signal evenly into multiple segments, a short-time harmonic model was established in each time period, and the signal was weighted for discrete Fourier transform with multiple orthogonal window functions. On this basis, a detection model was derived according with the F distribution. The model was designed to be independent of noise power, so the constant false alarm rate (CFAR) detection of signals in non-stationary noise environment could be carried out without statistical noise power before detection. The parameters that affect the performance of the algorithm were simulated and analyzed, and compared with the single window detection algorithm to verify the excellent detection performance of the algorithm in non-stationary noise environment.