Research On Detection Method Of Foreign Objects On PSD Based On Image Identification

With the rapid development of the transportation industry,national construction and the continuous increase of subway passenger flow,platform doors,as the main equipment for subway safety,have become one of the standard configurations of every subway.Because there is a certain gap between the platform door and the train,the gap makes it easy for passengers to rush to get on the train,causing passengers to be caught in the gap,items left in the gap,clothes and other items stuck in the gap,so there is Larger security risks.At present,the mainstream methods for detecting foreign objects include electric eye detection,automatic grating detection and infrared light curtain detection.These solutions have relatively large detection blind areas and blind areas of visual field,which cannot compare the safety,reliability,and real-time performance of trains.Good protection.To solve the above problems,this paper proposes a foreign object detection image recognition method based on multi-frame redundancy and multi-scheme redundancy.First of all,the principle of MagicAI and its key technologies such as MagicNet and MagicSegmentation are reviewed,and several detection models and detection algorithms are described.Based on the comparative analysis of actual effects,a YOLO model suitable for the scheme of this paper is selected.Secondly,using the MagicSegmentation technology of foreground detection Magic Net technology to perform real-time analysis of different principles and different angles on the video screen,to the greatest extent,ensure timely detection of danger,notify the PSD system and subway dispatch,so that the danger can be quickly stopped,so as to achieve a certain Area and space to achieve foreign body intrusion detection.Then,based on the above model,a foreign body detection system in the platform gap is realized.Finally,the system was deployed on the trial-production platform for testing,and the operation results verified the feasibility and effectiveness of the program.