Realizing automatic detection and tracking of targets is an important development direction of tank fire control system in the future. Firstly, with the method of transfer learning, we apply faster R-convolution neural network (Faster R-CNN) algorithm based on deep learning model to solve the detection problems of tank armored targets under complex background, and we achieve a higher detection accuracy compared with the traditional algorithm. Secondly, aiming at the shortcomings of existing tracking algorithms in the tank fire control system, we propose a composite tracking algorithm by combining Faster R-CNN with existing tracking algorithms to achieve automatic detection and stable tracking of the tank armored targets. Finally, we design an automatic detection and tracking system which uses dynamic scanning and staring imaging technique to realize fast and clear acquisition of battlefield images in a large field of view, and the algorithm of this paper is tested. Moreover, this paper also points out some problems that need to be solved when the deep learning algorithm is applied to the tank fire control system.