A UAV online aerial photography vehicle recognition and statistics method for congested road sections

Abstract

The invention discloses a UAV online aerial photography vehicle identification and statistics method for congested road sections. The method first constructs a congested road section aerial photography vehicle data set, then improves the original YOLOv3 network, and uses the K-Means++ algorithm to obtain The anchor size of the aerial vehicle data set of the congested road, and use the anchor size to train the real-time detection model of the aerial vehicle in the congested road based on the YOLOv3 convolutional neural network; The image is detected online in real time to identify the vehicle target. At the same time, the KM algorithm is used to establish an accurate matching relationship between the vehicle targets detected in two adjacent frames of images, and the number of new vehicles in the current frame compared with the previous frame can be accurately obtained, so as to realize the vehicle congestion in the congested road section. total. The vehicle identification and statistics method proposed by the invention has better flexibility, and realizes accurate, real-time and online statistics of the number of vehicles in the congested road section.

Description

technical field [0001] The invention belongs to the field of intelligent transportation, and in particular relates to a vehicle identification and statistics method for online aerial photography of unmanned aerial vehicles for congested road sections. Background technique [0002] With the continuous increase of my country's traffic mileage and the explosive growth of car ownership, coupled with the complexity of my country's road conditions and traffic conditions, traffic congestion and traffic accidents have occurred frequently. In order to solve the above problems, the research on ITS Intelligent Transportation System (ITS Intelligent Transportation System) has been widely concerned by scholars from all walks of life. Among them, vehicle identification and statistics are important contents of intelligent transportation research, especially for accurate detection and statistics of vehicles in congested road sections, which is helpful for timely and effective adoption of ta...

AI Technical Summary

Problems solved by technology

[0004] In addition, the existing vision-based vehicle recognition and statistical algorithms have low vehicle detection accuracy and poor real-time performance, and they are mainly used for detection and statistics of vehicles in a single frame of pictures, and the number of vehicles for continuous frames is relatively small. And often only the detected vehicle target center point distance is used as the discrimination standard for the same vehicle in consecutive frames, so that it is impossible to accurately distinguish the same vehicle in consecutive frames, which easily causes vehicle statistical errors

Images