Nowadays, huge volume of air quality data provides unprecedented opportunities for analyzing pollution. However, due to the high complexity, most traditional analytical methods focus on abstracting data, so these techniques discard the original structure and limit the understanding of the results. Visual analysis is a powerful technique for exploring unknown patterns since it retains the details of the original data and gives visual feedback to users. In this paper, we focus on air quality data and propose the AirInsight design, an interactive visual analytic system for recognizing, exploring, and summarizing regular patterns, as well as detecting, classifying, and interpreting abnormal cases. Based on the time-varying and multivariate features of air quality data, a dimension reduction method Composite Least Square Projection (CLSP) is proposed, which allows appreciating and interpreting the data patterns in the context of attributes. On the basis of the observed regular patterns, multiple abnormal cases are further detected, including the multivariate anomalies by the proposed Noise Hierarchical Clustering (NHC) method, abruptly changing timestamps by Time diversity (TD) indicator, and cities with unique patterns by the Geographical Surprise (GS) measure. Moreover, we combine TD and GS to group anomalies based on their underlying spatiotemporal correlations. AirInsight includes multiple coordinated views and rich interactive functions to provide contextual information from different aspects and facilitate a comprehensive understanding. In particular, a pair of glyphs are designed that provide a visual representation of the temporal variation in air quality conditions for a user-selected city. Experiments show that CLSP improves the accuracy of Least Square Projection (LSP) and that NHC has the ability to separate noises. Meanwhile, several case studies and task-based user evaluation demonstrate that our system is effective and practical for exploring and interpreting multivariate spatiotemporal patterns and anomalies in air quality data.

The propagation of pollutants between regions has become a noticeable factor affecting air pollution. Given the complicated propagation relationship, most of the existing works lack an effective perception mechanism of geographic correlations and time-varying features, which is crucial in exploring and understanding the propagation mechanism by integrating empirical knowledge and data inherent characteristics. In this paper, we abstract the complicated propagation relationship between regions as a dynamic network, and introduce visual analytics techniques to explore the spatiotemporal multivariate patterns of air pollution propagation. A particle tracking based model is first proposed to construct pollution propagation networks under multi-source factors. It combines numerical simulation and data characteristics simultaneously, and detects active pollution source areas based on long-term transport relationships and temporal correlations. Based on it, we extract propagation patterns and analyze the temporal evolution of diachronic propagation networks. Moreover, we design an interactive system to achieve an in-depth analysis of air pollution issues. Through elaborate multi-level glyphs and linkage views, the system facilitates users to perceive and explore propagation mechanism in spatiotemporal multivariate information, and compare propagation patterns from global and local perspectives. We present several case studies to demonstrate the usefulness of our work in air pollution propagation analysis.

Traffic vehicle detection in UAV aerial images facilitates traffic analysis and holds significant implications for traffic management and optimization. While drone technology provides unique aerial advantages for ground object detection, traffic vehicle detection in aerial images remains challenging. Small-scale, low-resolution vehicles in cluttered environments with frequent occlusions degrade detection accuracy, compounded by sensitivity to lighting variations during imaging. Moreover, current high-performance detection models demand excessive computational resources and parameters, conflicting with drones’ constrained edge computing capabilities and hindering practical deployment in traffic management systems. To address these challenges, this paper proposes a vehicle detection model based on feature optimization and precise convolution techniques. First, a plug-and-play feature optimization module is added to the backbone network’s output, using multi-branch attribute convolution to capture detailed local contextual features for small object representation. Second, the precise subsampling convolution module addresses information loss in standard convolution, enhancing fine-grained feature learning. Lastly, the micro precision perception pyramid attention network strengthens feature interactions between deep and shallow layers, improving detection precision for small objects and intricate details. Experimental results show that the proposed model significantly improves accuracy over baseline models, achieving 91.4% in detecting small objects while slightly reducing parameters. Additionally, a lightweight version of the model outperforms benchmark models in various metrics. This progress highlights the effectiveness of the strategy and showcases the model’s potential for real-world applications.