STROOBnet Optimization via GPU-Accelerated Proximal Recurrence Strategies

TL;DR

The paper tackles optimizing Spatiotemporal Ranged Observer-Observable Bipartite Networks (STROOBnets) for crime surveillance in New Orleans. It introduces a GPU-accelerated proximal recurrence method that combines spatial proximity and event recurrence via a bipartite distance matrix $DM$ and a radius threshold $r$ to improve observer centrality and coverage. Empirical results show proximal recurrence outperforms traditional clustering methods (K-means, DBSCAN) and mode clustering, enabling strategic insertion of new observers and yielding a more uniform degree centrality distribution. The approach scales to large datasets and generalizes to other observer-type networks, offering a practical framework for deploying surveillance assets in dynamic, high-crime environments.

Abstract

Spatiotemporal networks' observational capabilities are crucial for accurate data gathering and informed decisions across multiple sectors. This study focuses on the Spatiotemporal Ranged Observer-Observable Bipartite Network (STROOBnet), linking observational nodes (e.g., surveillance cameras) to events within defined geographical regions, enabling efficient monitoring. Using data from Real-Time Crime Camera (RTCC) systems and Calls for Service (CFS) in New Orleans, where RTCC combats rising crime amidst reduced police presence, we address the network's initial observational imbalances. Aiming for uniform observational efficacy, we propose the Proximal Recurrence approach. It outperformed traditional clustering methods like k-means and DBSCAN by offering holistic event frequency and spatial consideration, enhancing observational coverage.

Figures (17)

• Figure 1: Histogram of effectiveness scores (node degrees) for observer nodes in the initial STROOBnet, demonstrating a power-law distribution. The x-axis represents the degree count, signifying the ability of an observer node to detect events within a specified radius, while the y-axis shows the count of nodes for each degree.
• Figure 2: Heatmap of observer node effectiveness scores in the initial STROOBnet. Node spatial locations are plotted with color intensity indicating degree centrality and a gradient from blue (low effectiveness) to red (high effectiveness).
• Figure 3: Spatial distribution and observational coverage of events in the initial STROOBnet. Event locations are color-coded to indicate observational coverage: deep green for multiple observers, green for a single observer, varied colors for near-observers, and red for unobserved events.
• Figure 4: STROOBnet visualization post proximal recurrence integration. New nodes are represented in blue, directly observed nodes in green, and unobserved nodes in red, with edges indicating observational relationships.
• Figure 5: Histogram of node degree distribution for the new nodes introduced through the proximal recurrence strategy. Axes represent degree and node count, respectively.