Multivariate time series traffic dataset is usually large with multiple feature dimensions for long time duration under certain time intervals or sampling rates. In applications such as intelligent transportation systems, some machine learning methods being applied to traffic anomaly detections are computed under certain assumptions and require further improvements. Transport traffic time series data may also suffer from unbalanced number of training data where large amount of labelled training data available for a few popular classes, but with very small amount of labelled data for corner cases. In this paper, based on the recent long sequences prediction method Informer, an anomaly detection algorithm with an anomaly score generator is proposed that does not require any assumptions of data. The encoder-decoder architecture is adopted in the anomaly score generator. The encoder consists of three stacking ProbSparse self-attention mechanisms that significantly reduce computing complexity. The decoder incorporates two multihead attention layers and a fully connected layer to obtain an output of anomaly scores. Then a One-Class Support Vector Machines (OCSVM) is applied to be the anomaly classifier. The proposed algorithm is capable of detecting anomalies for both vehicle traffic flows and pedestrian flows. It has been verified by applying to a real-world dataset consisting of traffic flows recorded in 2021, as well as to a public anomaly detection dataset.
Accurate motion object detection (MOD) using in-vehicle cameras in driving vehicles is a challenging task. Several deep learning based motion segmentation approaches have been reported based on the interpretable optical flow feature. However, the interpretable optical flow feature has not been explored by object-level MOD approaches. In this paper, we propose a motion embedding pipeline (MEP) architecture that utilizes interpretable optical flow and deep learning to solve object-level MOD problems. The MEP is a three-stage pipeline that consists of an object detector, a novel feature extraction algorithm to capture relative motion between objects and the background, as well as a motion predictor for representation learning with stacked autoencoder to determine motions. A new dataset, Singapore motion object detection (SG-MOD) dataset is constructed in this work with much larger variations in urban environments. Experimental results show that the proposed MEP outperforms other pipeline-based architecture and deep learning based approaches on the SG-MOD and KITTI-MOD datasets in most metrics.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.