Abstract:Effective trajectory selection and classification are pivotal in user tracking systems utilizing spatiotemporal data collected from city sensors. However, the inherent limitations in sensor technologies and data collection point distributions often result in low-quality spatiotemporal data. Real-life trajectory classification encounters challenges due to the following: (1) high-order and sparse activity data encompassing both temporal and spatial contexts, and (2) inherent vagueness in the semantics of visited locations, making it difficult to represent behavioral intentions. Traditional statistics-based or trajectory-based feature approaches prove ineffective with non-discriminate features. In response to these challenges, we introduce a novel classification method that integrates fuzzy spatiotemporal features and crowd habit features. This approach involves feature extraction using the Time-Geo Hash (TGH) and User Transit Pattern and Similarity (UTPS) models, followed by the training of a machine learning classification model. On the basis of the performance indicators of classification models, we identify two classification algorithms, incorporate the Bagging algorithm from ensemble learning to enhance the UTPS classification model, and combine the TGH and UTPS models through specified rules. Extensive experiments demonstrate that our proposed model significantly outperforms other classification baselines when applied to a labeled real-life dataset, emphasizing its effectiveness in handling noisy and challenging spatiotemporal data for trajectory classification in user tracking systems.

A Cost Function Approach for Multi-Human Tracking

Beyond Traditional Driving Scenes: A Robotic-Centric Paradigm for 2D+3D Human Tracking Using Siamese Transformer Network

Multi-modal 3D Human Tracking for Robots in Complex Environment with Siamese Point-Video Transformer

Tracking Multiple People under Occlusion and Across Cameras Using Probabilistic Models

Multihuman Tracking Based on a Spatial–Temporal Appearance Match

Multi-person Multi-Camera Tracking for Live Stream Videos Based on Improved Motion Model and Matching Cascade

Multi-object tracking via discriminative appearance modeling.

Exploiting Pair-Wise Constraints Between Parts for Human Tracking

Multiple Human Tracking Based on Multi-view Upper-Body Detection and Discriminative Learning

MAT: Motion-Aware Multi-Object Tracking

Multi-Person Articulated Tracking With Spatial and Temporal Embeddings

Multiple Object Tracking by Multi-Feature Combination Based on Min-Cost Network Flow

3D Multi-Object Tracking in Point Clouds Based on Prediction Confidence-Guided Data Association

Human Motion Tracking by Multiple RGBD Cameras.

Multi-Person Pose Tracking With Sparse Key-Point Flow Estimation and Hierarchical Graph Distance Minimization

MotionTrack: Learning Robust Short-term and Long-term Motions for Multi-Object Tracking

Fuzzy Spatiotemporal Representation Model for Human Trajectory Classification

Tracking Multiple-person using Sparse Stereo Information

Multiple object tracking with appearance feature prediction and similarity fusion

Delving into the Trajectory Long-tail Distribution for Muti-object Tracking

A time-weighted metric for sets of trajectories to assess multi-object tracking algorithms