The invention discloses a motion feature and track feature fused target classification method, which comprises the following steps of: when a low-speed
small target is detected, automatically batching by a
radar to obtain track information of the target, visualizing to obtain motion features and track features, converting a motion
feature matrix and a track
feature matrix into fixed sizes, and classifying the motion
feature matrix and the track feature matrix; and sending the track features and the motion features with the
fixed length into an initial classification network constructed based on LSTM to generate an initial
classification result, and finally fusing the target classification results through an adaptive weighted summation strategy to obtain an accurate
classification result. According to the scheme,
radar original
point data is deconstructed into'motion feature sequences', the'motion feature sequences' are respectively input into the bidirectional LSTM network to extract the
time sequence dependency relationship, the dimensional difference between motion features and track features is eliminated through the feature space alignment technology, the physical significance consistency of fusion calculation is ensured, and the fusion calculation accuracy is improved. The recognition capability of maneuvering targets is improved, the robustness of the
algorithm is improved, and support can be further provided for classification of low, small and slow targets.