Video anomaly detection (VAD) focuses on identifying anomalies in videos. Supervised methods demand substantial in-domain training data and fail to deliver clear explanations for anomalies. In contrast, training-free methods leverage the knowledge reserves and language interactivity of large pre-trained models to detect anomalies. However, the current fixed-length temporal window sampling approaches struggle to accurately capture anomalies with varying temporal spans. Therefore, we propose VADTree that utilizes a Hierarchical Granularityaware Tree (HGTree) structure for flexible sampling in VAD. VADTree leverages the knowledge embedded in a pre-trained Generic Event Boundary Detection (GEBD) model to characterize potential anomaly event boundaries. Specifically, VADTree decomposes the video into generic event nodes based on boundary confidence, and performs adaptive coarse-fine hierarchical structuring and redundancy removal to construct the HGTree. Then, the multi-dimensional priors are injected into the visual language models (VLMs) to enhance the node-wise anomaly perception, and anomaly reasoning for generic event nodes is achieved via large language models (LLMs). Finally, an inter-cluster node correlation method is used to integrate the multi-granularity anomaly scores. Extensive experiments on three challenging datasets demonstrate that VADTree achieves state-of-the-art performance in training-free settings while drastically reducing the number of sampled video segments. The code will be available at https://github.com/wenlongli10/VADTree.
The automatic train operation(ATO)system for HXD1 locomotives has been incorporated in routine operational assess-ments on the Xi'an-Ankang line under the Xi'an railway bureau.This paper aims to address the challenges posed to the locomotive ATO system in future operational scenarios that involve crossing the railway bureaus and route boundaries.It presents an in-depth analysis of the difficulties faced by the ATO system in these scenarios including operation across multiple railway lines,track and yard transfers,and diverse train formations and vehicle types.The study focuses on the research and implementation of key technologies for operation cross-ing different bureaus and route boundaries,including electronic maps and operational scales,transfer-oriented operational planning and control,as well as operational planning and control for multiple formations and vehicle types.Field application results demonstrate that these technologies contribute to the safe operation of freight trains in operational scenarios involving different bureaus and route boundar-ies,by enhancing the adaptability of the locomotive ATO system.
The deceleration and stopping process of freight trains on downslope sections through air braking,may encounter low converted pressure values of the brake shoes per 100 tons of train weight.Failure to timely reduce train speeds below the braking limit and control train operation within the speed limit could potentially result in hazards such as overspeeding and overrunning.To address this challenge,a calculation method for train braking speed limits was designed,tailored for the locomotive automatic operation system.Based on the train braking speed limit curve,functions for train operation curve planning,driving control,and safety protection were de-veloped,which ensured the operational safety of trains in the automatic operation mode.Results from practical applications show the ef-fectiveness of the proposed approach in improving the efficiency of train operation compared with manual driving.
