A Vehicle Detection Method in UAV Aerial Images Based on Deep Learning

Abstract

The invention discloses a vehicle detection method in a UAV aerial image based on deep learning. First, collect aerial images of UAVs, and mark the vehicles in them to obtain the vehicle database; then, send the obtained vehicle database to the deep learning network for training until the deep learning network converges; finally, use the trained deep learning Network and weight files to detect vehicle objects in test images and output detection results. The invention has high precision and good robustness, and overcomes difficult problems such as environmental interference and illumination which are difficult to be solved by traditional image processing algorithms in the process of vehicle detection.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a vehicle detection method in aerial images of unmanned aerial vehicles. Background technique [0002] With the development of social economy, the transportation industry is developing rapidly, the number of vehicles is huge, and it is still increasing year by year, causing traffic accidents, vehicle congestion, vehicle chaos and other phenomena are more and more frequent. These traffic problems have seriously affected the daily travel of residents and increased the burden of ground traffic management. Although cameras are installed on key nodes of the city at present, this cannot visually display the traffic conditions of the entire road. Due to the portability and flexibility of UAVs, the use of UAVs for precise positioning and identification of vehicles has great advantages in detecting road traffic conditions. [0003] At present, vehicle detection algor...

AI Technical Summary

Problems solved by technology

However, this algorithm only combines some shallow features, and is easily affected by the image resolution, and the effectiveness of the algorithm is poor.

Images