Research On The Classification Of Airline Baggage Based On The Analysis Of 3D Morphological Characteristics

In order to achieve the goal of "simplified business and convenient travel" proposed by IATA regarding airline passenger service, many countries actively take actions in recent years, of which the self-service check-in of the passenger's baggage is one of the major tasks. In the support system where the self-service is achieved, a key technical issue needed to be resolved is to detect the shape of the passenger's baggage automatically. This topic studies the automatic classification detection of airline passenger's baggage, which mainly includes the acquisition of baggage depth point cloud images, the extraction of depth features, and the detection of baggage classification.First of all, it studies the method to acquire the cloud images, with deep information, at the airline baggage's place and issues related to features extracting. The features of baggage are extracted by analyzing airline baggage transporting rules, collecting the samples of the airline baggage and extracting prior knowledge, in combination with the analysis of the transportation rules and prior knowledge. The features of baggage show that the depth image can express the superficial information better. The depth images of airline baggage are obtained by depth image acquisition and pretreatment with active distance measuring sensor because the light is relatively dim and unstable in the channel of self-service baggage handling system.Secondly, it studies issues related to the classification based on the depth features of external shape of baggage. The classification of baggage roughly comes down to the functional curved surface. Baggage formed by the functional curved surface includes baggage of such categories as regular cube, sphere, cylinder, and cone; baggage formed by non-functional curved surface, including the shoulder bag, portable soft bag and other bag, is divided into baggage that may be checked and that may not be checked.Thirdly, with regard to the baggage formed by functional curved surface, it proposes to detect the algorithm based on the category of distribution features of depth value on the surface of airline baggage. The distribution features of point cloud depth value of baggage of different categories are different with the contour shape of point cloud on the horizontal projection; the baggage classification is detected with these two groups of features as the feature vectors. Regarding the baggage formed by the non-functional curved surface, it puts forward the adaptive clustering algorithm of point grids. Because the point cloud depth value distributes unevenly and change largely, the structure is discretized with grid segmentation method. The adaptive clustering algorithm is carried out according to the similarity of various grids and the distance between grids. Fitting the result of the adaptive clustering algorithm as the measurement of smoothness on the baggage surface, and the baggage classification is detected by the smoothness.Based on the above basic study, baggage samples of various categories are collected and placed on the conveyor belt in different ways; the point cloud information is acquired, and the classification is detected. The experiment validates the effectiveness and stability of deep image acquisition, features extracting and the airline baggage classification algorithm.