Research On UAV Ground Vehicle Target Detection Algorithm Based On Deep Learning

With the rapid development of China’s economy,China’s road mileage and car ownership have increased year by year,and various traffic problems have followed.Real-time monitoring of ground traffic and identification of the number and location of various types of vehicles on the road in a short period of time are of great significance for intelligent management of transportation networks,emergency response to accidents,and geological disaster rescue.The use of remote sensing satellites and aerial remote sensing are extremely susceptible to weather and cloud factors,so the use of drones to monitor ground vehicles based on deep learning has gradually become a trend.However,there are two problems with target detection for drone images: one is that the proportion of targets in the image is generally small due to flying height and other reasons;the second is that it is very easy for vehicles to block each other or vehicles are blocked by other objects Case.Both of these will adversely affect the detection.Therefore,this paper improves on the basis of the Faster R-CNN algorithm based on candidate regions.And experiment with it to check the detection effect of image target.The main research work of this article is as follows:First of all,for the problem of a large number of samples in deep learning,affine transformation is performed on the basis of the collected data set,which increases the number of samples and marks the data set at the same time.Experiments with some traditional target detection algorithms and deep learning based target detection algorithms to verify the advantages of deep learning based target detection algorithms.For the detection of UAV image targets,small and medium-sized targets are prone to misdetection and missed detection.A detection method based on improved feature image fusion based on Faster R-CNN is proposed.A redesign was carried out,and a fusion module was designed to fuse different levels of feature images to obtain multi-level feature information.Experiments show that the average accuracy of the improved algorithm(m AP)has reached 84.10%,which improves the detection accuracy and improves the detection ability of small and medium-sized targets.Although the feature image fusion UAV target detection algorithm has achieved certain results,it is still not ideal for the detection of occlusion objects.Therefore,the method of combining context information is adopted to realize the full use of the characteristic information;and the defects of the non-maximum suppression algorithm are improved.The experiment proves that the average accuracy(m AP)of the improved detection algorithm for object occlusion reaches 87.69%,which enhances the detection effect of the occlusion target.