Ensuring the efficiency and stability of the fusion process when processing the target echo data collected from the multi-station radar system is a core technical problem to be solved urgently in the field of signal processing. At present, fusion technology mainly focuses on the use of intuitive features such as amplitude correlation and spatial localization of echoes, but these features are relatively simple. However, due to the lack of comprehensiveness of manual feature extraction, this method is easy to lead to problems such as waste of signal resources, incomplete feature extraction, and insufficient generalization of discriminant algorithms. In order to solve this challenge, this paper innovatively proposes a jamming recognition strategy that integrates multi-radar cooperative detection and convolutional neural network to improve the performance of anti-spoofing jamming. By digging deep into the unknown information in the echo data, we extract multi-dimensional and deep feature differences, which surpasses a single spatial correlation and significantly improves the accuracy of interference discrimination.