Abstract: With the advancement of technology, the current electronic warfare landscape has become increasingly complex. Radar systems are confronted with electronic interference characterized by high coherence, strong deception, stealthiness and low power. This significantly degrades their detection and tracking capabilities, potentially rendering them combat ineffective. Therefore, the accurate identification of the types of active interference faced by radars is crucial for implementing targeted interference suppression. Lightweight convolutional neural networks (MobileNet), which can effectively capture spatial structural information in images without manual feature extraction, have exhibited excellent performance in image processing and classification. In this paper, a radar interference identification model is proposed based on MobileNet, which is validated by a dataset of time-frequency characteristics of radar active interference. Experimental results reveal that the established model achieves a high F1-score of approximately 0. 9 for radar interference identification and classification, outperforming models such as SIFT template matching and CNN in various metrics, thereby demonstrating superior classification performance.