| With the rise of artificial intelligence,computer vision has been applied to more and more fields,covering almost every aspect of daily life.Using deep learning algorithm to classify,locate and track the image information collected by camera has become a research hotspot in the field of computer vision.As a basic and important part of computer vision,object detection affects the effect of subsequent image understanding.Aerial vehicle detection can detect the location and type information of vehicles from aerial images,which can provide important information support for vehicle related application problems.However,in the actual detection process,it is often limited by hardware devices,and can not provide sufficient computing power.Generally,the deep learning object detection model with high accuracy is slow,while the simple network model with obvious speed advantage has poor accuracy.In order to meet the requirements of practical application,it is necessary to ensure the speed and accuracy of object detection algorithm.In order to meet the requirements of practical application,this paper proposes a vehicle object detection algorithm and a new up sampling structure in aerial images,which can give better consideration to both accuracy and speed.The specific research contents are as follow:(1)Yolov3-tiny is improved and the adjusted BiFPN structure is fused to get a new network structure.The network provides two-way access,which can enrich the semantic information of all level feature graphs.(2)Based on the nearest neighbor up sampling and deconvolution layer,a new up sampling structure is proposed to recover the detail information in the small-scale feature m AP and enhance the ability of the network to extract the detail information.(3)Experiments are carried out on different data sets to test the generality of the algorithm,and on this basis,a prototype system of aerial image object detection is designed,which can detect four kinds of objects. |
PDF Full Text Request