Passenger flow statistics method and system based on deep optical flow tracking

Abstract

The invention discloses a passenger flow statistics method and system based on deep optical flow tracking. The method comprises the steps of obtaining a pedestrian image; carrying out passenger flow statistics by using a preset passenger flow statistics model based on deep optical flow tracking, i.e., for any given pedestrian image, respectively outputting the central position of a pedestrian headarea, the radius of the pedestrian head area and the movement displacement of the pedestrian head area after forward operation of a deep neural network model; further obtaining the position of the pedestrian head area of the current frame and the identity address relative to the previous frame, and updating the tracking list; and finishing passenger flow statistics by adopting a passenger flow statistics method. According to the invention, deep learning is adopted to realize position detection of a pedestrian head area and tracking of the pedestrian head area in parallel based on a deep neural network model; multi-scale semantic information and inter-frame optical flow information of the images are comprehensively utilized, passenger flow statistics is more accurate, and robustness is higher; the deep neural network model is designed based on the strategy of anchor-free, and the system operation speed is higher.

Description

technical field [0001] The invention relates to the technical field of pedestrian detection and tracking and passenger flow statistics, in particular to a method and system for passenger flow statistics based on deep optical flow tracking. Background technique [0002] Since passenger flow is an important basis for measuring the busyness of public places, accurate real-time passenger flow information has important guiding significance for managing public places and scheduling public facilities. For example: for the bus transportation system, by mastering the passenger flow at each station and time period, the bus company can adjust the route and increase or decrease vehicles more reasonably, which improves the utilization rate of resources. For shopping malls and supermarkets, accurate grasp of real-time passenger flow can dynamically adjust the number of staff and formulate reasonable marketing strategies. [0003] At present, passenger flow statistics mainly include the f...

AI Technical Summary

Problems solved by technology

Although this type of technology is low in cost and simple to implement, when there are dense and crowded crowds, the accuracy rate will seriously drop or even fail.

[0005] (2) Statistical techniques based on video image processing, such as "A Passenger Flow Statistics Method Based on Adaptive Template Matching" (Chinese Patent Application No.: CN201010509217), which is a typical two-dimensional image detection and tracking technology. The technology can achieve good statistical results in some specific scenes, such as indoor entrances with stable light, but in outdoors with unstable light, such as on a bus, this method will have serious errors

This type of technology can locate the position of people more accurately, and it also has a better effect on complex environments and crowded crowds, but the algorithm is more complicated, the operation efficiency is low, and the hardware cost is also high.

Images