Normalized convolutional network radar target detection method based on pre-statistical processing

The invention belongs to the technical field of radar signal processing, and provides a normalized convolutional network radar target detection method based on pre-statistical processing. According to the method, firstly, candidate position information of a target trace point is given by adopting low-threshold or normal-threshold CFAR detection according to needs, and then radar video image slices near the trace point are manufactured by taking the trace point as a center; after image slices are subjected to improved normalization processing, the image slices are sent to the trained dichotomous convolutional neural network for target type judgment, real target trace points are reserved, false target trace points are eliminated, automatic feature extraction and low false alarm radar target detection with strong generalization ability are realized, and the target detection efficiency is improved. The efficiency of feature extraction is improved, and a certain generalization ability is maintained. 本发明属于雷达信号处理技术领