| Railway freight transportation is of great significance in our daily lives.It is one of the main modes of modern transportation and also one of the two basic modes of transportation that constitute land freight transportation.Therefore,safety inspection plays a crucial role in the process of railway freight transportation,as it can ensure the safety and integrity of rails,trucks,and goods during the railway freight transportation process.However,there are also many problems with railway freight,such as unclosed cabin doors.However,traditional security detection methods use manual detection methods,relying on the experience and status of managers.If situations such as manager fatigue occur,it may lead to misjudgment.With the gradual maturity of artificial intelligence technology,it has been widely applied in various aspects.Based on this,we have designed a three-dimensional detection system for railway freight yards,which is used to detect three safety hazards: whether the cabin door is safely closed,whether the carriage is damaged,and whether there are obstacles on the track when the freight train departs.The 3D detection system for railway freight yards consists of two parts: a 3D detection algorithm and a 3D detection system.The 3D detection algorithm part implements a 3D point cloud object detection algorithm for detecting safety hazards of trucks during departure through Tensorflow,and then uses the 3D point cloud adversarial attack algorithm to ensure the robustness of the model while expanding the dataset.The 3D detection system consists of three parts: an embedded front-end,a server back-end,and a web frontend.The main function of the embedded front-end is to use Kinect v2 cameras to collect3 D point cloud data and store it? The main function of the server backend is to receive3 D point cloud data from the embedded front-end for detection and storage,while also receiving uploaded data from the interface front-end for detection and storage? The main function of the web front-end is to visualize the 3D point cloud data and security detection results of railway freight yards,and provide an available railway freight yard security hazard detection platform for railway freight yard managers.This thesis focuses on the safety hazards of freight cars in railway freight yards during departure,and designs a three-dimensional detection system for railways.The main achievements and innovative points are as follows:(1)Aiming at three potential safety hazards in the railway freight yard,such as whether the hatch is closed safely,whether the carriage is damaged,and whether the track has obstacles,a 3D point cloud target detection algorithm is studied.This algorithm can achieve a high average accuracy of 87.52%,and can achieve a high detection efficiency of 10.33 FPS,which can basically meet the needs of the detection of potential safety hazards in the railway freight yard.(2)Aiming at the target detection algorithm,a 3D point cloud anti-attack algorithm is studied,which increases the data set samples while ensuring the robustness of the model.(3)A 3D detection system for railway freight yard is designed and implemented.The system meets the functional requirements of detecting safe departure in railway freight yard,and has the characteristics of safety,scalability and portability. |
