The invention relates to a fast pedestrian detection method based on millimeter wave radar and visual fusion. The region of interest is scanned by millimeter wave radar, the received signal is extracted, and the received signal in the same frame is condensed and processed. Two buffers for storing valid target signals and abnormal signals are established respectively; The effective target center istransformed from radar coordinate system to image coordinate system to obtain the projection point of the effective target signal on the image. For the training of similar object detection, the INRIApedestrian data training set is used as a positive sample, and the window is randomly sampled in the image without pedestrians as a negative sample to participate in the training. The target area isscanned by the selected window, and pedestrian targets are detected by the similarity training results. If pedestrian targets are found, the scanning is stopped. Otherwise, the window moves until a pedestrian target is found. The invention avoids the low detection efficiency caused by the blind search of the traditional sliding window algorithm, and achieves the purpose of accelerating pedestriandetection.