This paper studies the problem of target detection in UAV aerial images using deep learning technology. Let''s start with single shot multibox detector target detection algorithm, and then improve it. Dense feature extraction network is used to improve the feature extraction ability of the algorithm, so as to improve the detection accuracy of the algorithm. Aiming at the problem of network real-time, packet convolution is introduced into the algorithm, which greatly reduces the amount of network parameters and improves the speed of network reasoning. In order to solve the positive and negative problems in training for the sample imbalance problem, this paper improves the loss function of the original algorithm and uses focal loss to improve the original loss function, which further improves the convergence speed and accuracy of the network. Finally, the superiority of the algorithm in target detection accuracy and speed is verified by simulation.