Special issue on intelligent systems and applications in vision, image and information computing

Albert Dipanda, Ernesto Damiani, Kokou YétongnonJ. Reliab. Intell. Environ 2016 · DOI

This special issue of the Journal of Reliable Intelligent Environments (JRIE) presents a curated selection of expanded research originally debuted at the SITIS 2015 conference in Bangkok. The collection focuses on bridging the gap between theoretical intelligent systems and their practical, "reliable" deployment in real-world environments, specifically addressing challenges in vision, image processing, and distributed information computing.

Addressing the Challenges of Intelligent Environments

As intelligent systems transition from controlled laboratory settings to complex "intelligent environments"—such as smart cities, offshore cloud networks, and wearable ecosystems—they face significant hurdles in dependability and scalability. This editorial highlights four primary areas of concern:

  1. Network Instability: Maintaining stable business communication when offshore cloud access is plagued by intentional blocks or high latency.
  2. IoT Fragmentation: The lack of unified, forward-compatible frameworks that prevent device obsolescence as protocols evolve.
  3. Vision Complexity: The high computational cost and environmental sensitivity of detecting faces in diverse imagery, including cartoons and real-world photos with occlusions.
  4. Biometric Scaling: The difficulty of performing accurate 1:N fingerprint matching at high speeds as databases grow to massive scales.

Methodological Innovations

The selected papers propose novel solutions to these persistent problems:

Proactive Network Virtualization

Fujikawa et al. utilize RTT (Round Trip Time) differential calculus to proactively detect the start of network restrictions. Unlike reactive systems that wait for a connection to fail, this system automatically switches to a VPN to bypass latency and reverts to the standard Internet once the restriction ends. This tool was validated through long-term usage across more than 50 international offices.

The "One IoT" Framework

Banda et al. propose a reference architecture and a minimal criteria protocol designed to be unconditionally forward compatible. This "One IoT" philosophy allows manufacturers (OEMs) to generate user interfaces on the fly based on device functionality, ensuring interoperability is a baseline requirement rather than an afterthought.

Hybrid Skin Color Modeling

Yadav and Nain address vision complexity by combining multiple color spaces—RGB, YCbCr, and HSV—for localized skin segmentation. By following this with morphological analysis (searching for eye-mouth holes and calculating eccentricity ratios), the system successfully reduces false positives in both real-human and cartoon images, even in the presence of noise, beards, and glasses.

Geometric Fingerprint Indexing

Prasad and Jain introduce two feature extraction schemes: neighboring relation using rectangles and triangle spirals. These features are organized using dynamic clustering and m-ary trees. This geometric approach achieved a 100% hit rate on tested databases, significantly speeding up the search process in high-dimensional biometric databases.

Key Insights and Results

The research presented in this issue underscores several critical shifts in the field:

  • Proactive vs. Reactive: Using differential calculus on RTT allows systems to "sense" incoming restrictions and pivot before quality-of-service drops.
  • Multi-Model Fusion: Combining three different color models in computer vision significantly reduces false-positive rates compared to single-model approaches.
  • Efficiency in Scaling: Moving from simple template matching to geometric constructs like triangle spirals allows for much more efficient indexing in large-scale systems.

Future Outlook

While these papers provide robust solutions, the guest editors acknowledge that investigating intelligent systems in image and information computing remains one of the most challenging tasks in computer science. Future work must focus on broader industry adoption of IoT standards and further refining how these systems handle quality-related issues in increasingly complex, large-scale information environments.

Summary updated

BibTeX
@article{dipanda2016special,
  author = {Albert Dipanda and Ernesto Damiani and Kokou Yétongnon},
  title = {{Special issue on intelligent systems and applications in vision, image and information computing}},
  journal = {J. Reliab. Intell. Environ},
  year = {2016},
  doi = {10.1007/s40860-016-0032-8},
}
Figure