To address the image distortion caused by the large-field camera in the panoramic system and the undulation and shaking of the stitched image of the target caused by the motion or sway of the mounting base, based on the distortion principle and the dynamic stabilization mathematical model for panoramic images, the distortion correction is performed for origin images captured by the camera.Then, in order to obtain a panoramic image, rotation transformation, spatial projection, and smoothing of each pixel of the original image are obtained according to the posture data of the installation base. Finally, algorithms based on deep learning are used for target detection and recognition. The results of training and testing on the established data set show that the algorithm designed in this paper can better correct the panoramic image, and the real-time detection speed of ship target is up to 25 fps, and the accuracy rate is up to 80%, which shows that the algorithm can meet the requirements of image stability and real-time detection of ship target under dynamic conditions.