Recently, the research of aerial images, images and videos taken by UAV(unmanned aerial vehicles) have developed rapidly. People use these images to get useful information. For example, investigate the density of vehicles, analyze the development of urban suburbs. Symmetry as one of the important features of objects, has long been a research hot area in computer vision. Symmetry detection aims to extract symmetric information from given objects. Aerial images and unmanned images have their special view angle. The symmetrical characteristics of industrial design leads to the special nature of the image from the top view angle. It shows more symmetry of the object more. The objects always like airplanes, ships, automobiles or industrial products like oil drums and chimneys. An aerial object detection method based on improved YOLOv3 algorithm is proposed in this paper. First, a symmetry prior is introduced into the proposed network based on the properties of symmetry patterns. For new convolutional layer, symmetric constrains are added during the process of kernel weights update. The symmetric kernels help the proposed network to find the corresponding symmetry objects. In addition, this paper uses k-means to calculate the size of anchors in order to make the size more suitable for the size of the aerial image object. Compared with the original YOLOv3 algorithm, the improved YOLOv3 algorithm can effectively solve the missing detection of symmetry objects and improve the confidence of detection.