Image Energy Pilot, FA Soil-Salt Recognition Approach(Neuro-based Recognition and Control for Alife(OS),Session: TA1-A)

Xin Wang, Masanori Sugisatka, Wenli XuThe Abstracts of the International Conference on Advanced Mechatronics Toward Evolutionary Fusion of IT and Mechatronics ICAM 2004 · DOI

Abstract

Reliable long-distance navigation for mobile robots remains a significant challenge due to factors such as surface slippage and sensor inaccuracies. This research explores the development of a vision system for "Alife" robots intended for service roles in environments like hospitals and homes. The proposed approach utilizes an "image energy" pilot and a hierarchical classification method to recognize complex textures, specifically focusing on soil-salt recognition in industrial or service contexts.

Introduction

In both industrial inspection and domestic service environments, robots encounter visual scenes characterized by a chaotic mix of objects with varying textures and gray-value gradients. These complexities often hinder standard computer vision systems from achieving accurate classification. This paper addresses the gap between raw visual data and reliable object recognition by proposing a hierarchical classification framework that leverages "image energy" and auto-relation functions.

System architecture for the Alife robot vision system, illustrating the flow from image capture to classification.

Methodology

The proposed vision system operates through a two-stage process designed to handle noisy visual environments effectively.

1. Pre-processing

To mitigate the impact of environmental noise while maintaining structural integrity, a traditional rank-value median filter is applied to the raw input images. This step is crucial for preserving important edges that define object boundaries amidst complex textures.

2. Feature Extraction and Classification

The core innovation lies in applying an auto-relation function to "image energy." In this context, image energy likely refers to the distribution of local variance or specific frequency components within the image. By analyzing these energy patterns, the system constructs a "piloting set"—a reference collection of elements used for hierarchical classification. This allows the robot to process visual information in a "coarse-to-fine" manner, narrowing down object identities through successive layers of detail.

An example of the textured, high-gradient images that the system must process to identify specific objectives.

The Piloting Set Concept

The "piloting set" serves as a structured framework for recognition. Rather than attempting a single-step identification, the system uses the piloting set to categorize visual inputs into broad groups before refining the search for specific details. This iterative approach is particularly effective at handling gradients in texture and gray values, treating these variations as informative clues rather than mere noise.

Conclusion and Future Work

While the theoretical framework for the image energy pilot and piloting set is established, this research is in its early stages. The authors indicate that further investigation is required to provide quantitative benchmarks and to test the system's robustness across diverse industrial and service scenarios. Future reports will focus on the empirical results of this hierarchical recognition approach in real-world Alife robot applications.

Summary updated

Abstract

For a mobite robot, it is difficu}t to realize long distance navigation reliab]y because of the slipper, the errors of the sensors, etc. In view of above problems, we explored an Alife robot used for the service such as in hospitals and home. In this rebot, we have deyeloped the vision system,
BibTeX
@inproceedings{wang2004image,
  author = {Xin Wang and Masanori Sugisatka and Wenli Xu},
  title = {{Image Energy Pilot, FA Soil-Salt Recognition Approach(Neuro-based Recognition and Control for Alife(OS),Session: TA1-A)}},
  booktitle = {The Abstracts of the International Conference on Advanced Mechatronics Toward Evolutionary Fusion of IT and Mechatronics ICAM},
  year = {2004},
  doi = {10.1299/jsmeicam.2004.4.39\_2},
}
Figure