Integration of GIS and Moving Objects in Surveillance Video

Yujia Xie, Meizhen Wang, Xuejun Liu, Yiguang WuISPRS Int. J. Geo Inf 2017 · DOI

TL;DR

This paper presents a novel framework for integrating Geographic Information Systems (GIS) with surveillance video data to overcome the limitations of traditional systems. By extracting, georeferencing, and fusing moving object information with 3D GIS models, the system enables more efficient searching, analysis, and visualization of dynamic events in a spatial context.

What Problem Does This Paper Address?

Modern surveillance systems generate vast amounts of video data, leading to challenges in efficient searching, redundant data transmission, and the inability to spatially contextualize events. Traditional video surveillance systems struggle with analyzing the spatial-temporal behavior of moving objects and lack the ability to precisely locate these events within a geographic framework. This paper aims to bridge this gap by integrating GIS with surveillance video to create a more intelligent and context-aware monitoring platform.

Methods

The core of this work lies in the proposed integration model and the GIS–MOV system architecture. Key techniques include:

  • Moving Object Detection and Tracking: Standard methods like background difference and inter-frame difference are used for detection, followed by tracking algorithms to extract attributes like speed, position, and trajectory.
  • Geo-spatialization of Video: This crucial step involves establishing a mapping relationship between image coordinates and geographic coordinates. The paper utilizes a homography matrix method for this purpose, requiring at least three pairs of corresponding points to solve for the transformation matrix.
  • Fusion Patterns: The paper defines several fusion patterns for integrating GIS and video data:
  • Image Projection: Video images are projected as textures onto a geographic scene model.
  • Object Projection: This is further divided into:
  • Foreground and Background Independent Projection: Moving object sub-images are projected onto their spatial-temporal positions, while the background is mapped as a texture.
  • Foreground Projection: Only moving object sub-images are projected, omitting the background.
  • Abstract of Foreground Projection: Moving objects are represented by semantic icons.
  • GIS–MOV System Architecture: A service-oriented architecture is proposed, comprising a Function layer (data processing), Data layer (storage), Service layer (data services), Business layer (service selection), and Representation layer (user interface).

Datasets

Not explicitly discussed in the paper. The authors mention using "a set of video images" for testing compression efficiency, but no specific datasets or benchmarks are named.

Key Insights and Ideas

  • Spatial Context is Key: Simply analyzing video data in isolation is insufficient. Integrating it with GIS provides a critical spatial and temporal context that significantly enhances understanding and analysis.
  • Multiple Fusion Strategies: The paper highlights that no single fusion method is optimal for all scenarios. Offering multiple patterns (foreground/background, object-only, semantic icons) allows users to choose the representation that best suits their needs, balancing detail with compression.
  • Moving Objects as Structured Data: While raw video is unstructured, extracted moving objects and their trajectories can be treated as structured data, making them amenable to GIS integration and analysis.
  • Compression Benefits: The proposed integration, particularly with abstract foreground projection, leads to significant data compression, addressing the issue of massive video data volume.

Key Results

  • Improved Visualization: The GIS–MOV system allows for the visualization of moving objects within a 3D geographic scene, offering a more intuitive understanding of events compared to traditional video interfaces.
  • Data Compression: The paper demonstrates significant data compression rates (on the order of 10⁻³ for the first two layers and 10⁻⁵ for the third layer) through its hierarchical compression approach, effectively reducing storage and transmission requirements.
  • Enhanced Analysis: The system enables applications like trajectory clustering and video synopsis on a geographic scene, providing users with a clear understanding of general dynamic trends and summarized event information.
  • Advantages over Existing Methods: Compared to video-augmented GIS, the proposed method offers a more integrated representation. Compared to previous GIS and surveillance video integrations, it provides a clearer understanding of dynamic video information in a geographic space through various fusion patterns.

Limitations and Future Work

  • Limitations:
    • Video Information Loss: Depending on the fusion pattern, some methods may omit background information or object sub-images, leading to potential loss of detail.
  • Complex Dynamic Scenarios: The system may struggle to represent highly complex dynamic video information, such as scenes with a large number of moving people or vehicles.
  • Single Camera Focus: The current study primarily focuses on video data acquired from a single camera.
  • Future Work:
    • Multi-Camera Integration: Extending the system to integrate GIS and moving objects extracted from multiple cameras in a network.
  • Moving Camera Integration: Incorporating video data from moving cameras.
  • Camera-Network Integration: Further research into integrating GIS and moving objects from camera networks with multiple moving cameras.

Summary updated

BibTeX
@article{xie2017integration,
  author = {Yujia Xie and Meizhen Wang and Xuejun Liu and Yiguang Wu},
  title = {{Integration of GIS and Moving Objects in Surveillance Video}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2017},
  doi = {10.3390/ijgi6040094},
}
Figure