Research On Key Technologies Of Airport Self-service Baggage Based On Depth Camera

With the development of society,the demand for passenger travel is increasingly strong,and how to improve the travel efficiency of the airport is a subject of continuous research at home and abroad.Airport baggage self-check service is a service that can greatly improve travel efficiency and save travel time.The key technology is the automatic and accurate detection of the shape of the baggage that passengers need to check.The subject mainly studies the automatic classification and detection of luggage in a wide field of view,involving the research of 3D point cloud acquisition and processing,point cloud registration,appearance detection and luggage classification detection algorithms.Firstly,this article outlines the development of airport baggage self-checking technology and solutions,analyzes the shortcomings of the existing solutions,and proposes a machine vision solution for baggage detection based on depth cameras.This solution can save a lot of labor costs and the shortage of manual detection,and has high practical application value.Secondly,the needs of baggage detection are analyzed,and the overall technical solution of the detection system is described in detail,including the optical platform,software design and algorithm design.Starting from the actual application scenario,for the height of the self-check-in checked baggage system can not be too high and meet the requirements of a larger field of view,a solution is proposed that uses two depth cameras to extract the 3D point cloud information of the luggage and uses 3D point cloud automatic registration.Next,the principle and process of 3D point cloud mosaic algorithm are analyzed in detail.The 3D point cloud registration technology used in this paper includes 3D-SIFT key point detection,FPFH feature solution,k-means feature point matching and other initial registration techniques and NDT precise registration techniques.In addition,the accelerated version of NDT accurate registration technology is used to improve the efficiency of the stitching algorithm.Then,based on the acquired point cloud data and depth images,starting from the morphology and depth information of the image,using machine learning and other technologies to study the actual scene of baggage check.Corresponding algorithms are designed to identify multiple pieces of luggage,soft and hard bags,and luggage size,respectively.After all the items have been tested,determine whether the baggage meets the shipping standards.Finally,based on the above basic theoretical research,a large number of different luggage items were collected as samples,and they were placed on the luggage conveyor in different positions to simulate the actual detection scene,and then the point cloud information was collected to verify the classification detection algorithm.A large number of experiments have verified the feasibility of the project scheme,and the effectiveness and stability of the detection algorithm.The baggage detection scheme proposed in this paper has an accuracy rate of 95%,a soft and hard bag recognition rate of 93.5%,a multibaggage recognition rate of 85.6%,and a size detection false positive rate of 6%.The average test process takes about 30 s to 40 s.This article studies the self-service baggage check system from three aspects: hardware,software,and algorithms.The solution used has not been reported in China.The research on the subject of self-service baggage check system has high reference value.