Abstract:

According to the international standard of baggage consignment methods for airline passengers, we study the classification method based on the variance of the surface depth of the baggage, and attempt to provide research basis for consignment methods. The depth image above the baggage conveyor belt is acquired by the sensor of Kinect, and pixel values of the baggage area are extracted so as to calculate their variance to classify the baggage roughly. In combination with prior knowledge of baggage’s threedimensional surface morphology, and according to the distance between grids and the variance of the depth value’s difference, we design a selfadaptive clustering algorithm based on grid similarity, which fits the clustering results of the area of each top unit as well as the number of all the top units. The threedimensional morphological characteristics of baggage are analyzed and the classification of baggage is determined. Experimental results indicate that this algorithm is of low complexity, and the classification is accurate and effective. 

Key words: classification, airline baggage, mean square error, selfadaptive clustering, 3D morphology