Method for counting vehicles in lane based on full convolutional network

Abstract

The invention discloses a method for counting vehicles in lanes based on a full convolutional network, which comprises the following steps of: s1, setting a monitoring detection area, extracting videoframe information, performing lane division processing on the video frame information, and setting an interested area in each lane area, the interested area being an area needing to be monitored anddetected in each lane; s2, processing the video frame information to obtain a vehicle target of the current monitoring image, and outputting a vehicle density map; s3, according to the vehicle densitymap in the step S2 and the lane area in the step S1, obtaining the number of vehicles in any lane; s4, presetting a threshold value yi for the number Ni of vehicles which can be accommodated in the lane i; if Ni is greater than yi at any time within the time T, starting speed judgment; s5, constructing 3D convolution, and obtaining the average speed vi of vehicles in the lane i within a short time t; and S6, judging whether the lane i is congested or not by combining the step S3 and the step S5. Video frames can be processed efficiently in real time, vehicle counting is carried out, and the circulation condition of a road is judged.

Description

technical field [0001] The invention relates to a method for counting vehicles, in particular to a method for counting vehicles in a lane based on a full convolutional network. Background technique [0002] With the development of the economy, the rapid increase of household vehicles, traffic congestion and other problems emerge in endlessly. In order to solve these traffic problems, intelligent traffic control means came into being. The intelligent transportation system is a system that comprehensively utilizes computer technology, image processing and intelligent detection technologies, and the detection of video traffic flow is an important part of the intelligent transportation system, which can provide a reliable basis for traffic planning and management. [0003] At present, the commonly used video traffic flow detection method is an adaptive virtual coil traffic flow detection algorithm. The detection method is based on the image binarization principle, performs sec...

AI Technical Summary

Problems solved by technology

The technical solution is to use the first needle image to initialize the background model. Due to the complex outdoor environment, the camera often shakes or shifts. At this time, it will interfere with the detection of the foreground, thereby increasing the missed detection and false detection of the vehicle. Rate

When the weather changes, the foreground detection will also have a large error

Images