Rail transit passenger flow density estimation system and method based on target detection

• Main Authors: ZHAO HUAWEI, LU SAIJIE, LIU WEIWEI, LIU GUANGJIE, PAN SHANGKAO
• Format: Patent
• Language: Chinese; English
• Published: 09.04.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC COMMUNICATION TECHNIQUE; ELECTRICITY; HANDLING RECORD CARRIERS; PHYSICS; PICTORIAL COMMUNICATION, e.g. TELEVISION; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a rail transit passenger flow density estimation system and a rail transit passenger flow density estimation method based on target detection, which are realized on the basis of an Atlas 200 development board, and are implemented by the following steps of: acquiring video data to perform hardware decoding, acquiring a picture in a YUV format, and preprocessing the picture, including size adjustment and format conversion; loading the trained crowd counting model, and completing an inference process on the preprocessed picture to obtain a feature vector of the picture; processing the feature vector output by model reasoning to obtain the number of pedestrians in the picture, and comparing the number with the rated capacity to obtain a current density value; based on data collected by the video collection module, testing the model by using a MindStudio tool to obtain the precision of the model on a test set; if the precision is reduced compared with that of a GPU platform and the result i