Abstract: Aiming at the limitation, of the traditional unmanned aerial vehicle detection method relying on a single feature, this paper proposes a feature fusion convolutional neural network structure based on passive radar, and the depth learning method is used to realize UAV detection. In this method, the digital terrestrial multimedia broadcast signal is adopted as the external illuminator signal, and the short-time Fourier transform time-frequency map and the wavelet transform time-frequency map are respectively constructed by receiving its echo signal. These two types of time-frequency maps are input as the dataset to fuse a convolutional neural network. After the features are extracted through the calculation of the network-related layer, the features are fused by the Concat layer, Finally, the detection results are obtained in the output layer. The experimental results show that under −10 dB, the accuracy of UAV detection using dual feature fusion structures reaches 83%, and under other different signal-to-noise ratios, the accuracy of UAV detection using two feature fusion structures is significantly superior to that of single-feature detection methods, which verifies the effectiveness of the proposed model.