A Method Of Object Positioning Based On Monocular Vision And Its Application In Crowd Evacuation Path Planning

Evacuation of people is an important issue involving public safety.In recent years,with the social and economic development,the process of urbanization has continued to advance.The good living conditions and employment opportunities in cities have attracted a large number of external people to work and live in the cities,which has led to a rapid increase in urban population.In order to provide more living and living space for people living and working in the city,more and more large buildings are beginning to appear in the city.These large buildings provide space for people to live and live in the city,but the large population contained inside hides huge security risks.When an emergency situation occurs,how to evacuate the internal people scientifically and efficiently becomes a problem that must be paid attention to and solved.The related research on crowd evacuation originated around the 1970 s,and today it has nearly 50 years of research history.However,crowd evacuation algorithms are mostly concentrated and explore the optimal path.In actual applications,it lacks real information such as pedestrians and the environment.Makes many excellent algorithms difficult to apply.In the direction of crowd movement modeling,many scholars have proposed many data-driven crowd movement models in order to improve the performance of modeling simulation.However,they still face problems such as lack of real crowd movement data and complex extraction of pedestrian movement information.There is a lack of crowd evacuation data for real scenes,and there is a problem of lack of comparative data in the evaluation of crowd movement simulation modeling.In view of the above problems,we propose a method for obtaining pedestrian information based on monocular vision,and combine it with an evacuation algorithm.The main work and innovations of this article are as follows:(1)A pedestrian positioning method based on monocular vision is proposed.In this paper,we propose a pedestrian positioning method that combines target detection and camera calibration.Our method first detects a pedestrian in an image using an object detection algorithm and obtains its pixel coordinates.The camera calibration part is calibrated by Zhang 's calibration.Method to establish the conversion relationship between pixel coordinates and world coordinates.Finally,we convert the pedestrian pixel coordinates obtained by target detection to coordinate coordinates and transform them into their world coordinates to achieve the positioning of pedestrians.(2)The sort multi-target tracking algorithm is optimized.The article mainly improves the target tracking algorithm from two aspects.1.Based on the non-linear characteristics ofpedestrian movement,optimize the state prediction algorithm,and improve the accuracy of pedestrian state prediction by replacing the Kalman filter algorithm with the unscented Kalman filter.2.Improve the matching method of the Hungarian matching algorithm in sort,add motion change difference value matching to distance-based matching,and consider the similarity between pedestrian dynamic information based on the observation of static distances,thereby improving the accuracy of matching.Sex.(3)A pedestrian counting method based on clustering and adjustable counting domain is proposed.This method improves the accuracy of pedestrian counting by clustering and adjusting the counting domain.This method considers pedestrians that tend to aggregate as a whole,and clusters pedestrians with a distance less than a certain threshold into a cluster by clustering,and records the aggregated pedestrians.The number of people,when a group of people pass through an exit,only needs to count the flow of people according to the aggregated number of clusters,without the need to perform high-precision detection of pedestrians in the occluded part.The use of adjustable counting fields can effectively prevent pedestrians from being unable to observe the crossing of the counting line due to congestion.The above two methods can effectively improve the accuracy of pedestrian flow counting.At the end of the article,we combined real data with the evacuation algorithm.The experimental results show that the proposed pedestrian positioning method has good accuracy,and the pedestrian data can be well combined with the path planning algorithm.The question is significant.