IRIS Research Assistant

 Chang Cheng
Ph.D. Student

 Examples of images that is recognized by our algorithm as originating from the same place in a large outdoor environment.

 We developed an object-based place recognition and loop closing method. Instead of directly using large numbers of SIFT features as visual landmarks, we first use a novel image segmentation algorithm to segment the input scene image into regions that may correspond to objects or parts of objects. Based on these image regions, we further detect a set of landmark objects to represent a place and only those SIFT descriptors that were contained in these landmark objects were retained in a database. We also designed a range tree data structure to organize these landmark objects to increase the matching efficiency. Experiments show that place recognition can be achieved accurately and efficiently with these landmark objects in a large outdoor environment.

Examples of image segmentation results

Image segmentation for natural scene images

 We propose a novel scene image segmentation algorithm based on Perceptual Organization. Different from existing methods, our method detects object boundaries mainly based on the shape, size and spatial relations of image regions. We developed a Perceptual Organization model that can capture the non-accidental structural relations among a group of object parts by quantitatively incorporating various Gestalt laws together. By doing this, the output of the Perceptual Organization model is close to a human’s perception. To our knowledge, this is the first work that systematically applies Gestalt laws in image segmentations. Our experimental results show that our proposed method outperformed three competing image segmentation approaches and achieved good segmentation quality on various natural scene environments.

.          

Example of Perception Organization: Most people tend to group a and b together.

SafeBot autonomous mobile platform

The purpose of this project is to build an autonomous mobile robotic system. Our SafeBot mobile robot consists of an intelligent system, a range sensor brick and a low-profile mobile platform. The intelligent system runs in a remote computer. The intelligent system receives the range sensor reading and makes motion plan accordingly. The range sensor brick is mounted on the low-profile mobile platform. The range sensor brick is designed to be a self-sufficient system. It is composed of sensor, communication, intelligence, and power subsystems. This modular design allows it to run independently. The low-profile mobile platform consists of two independently-powered tracks. We designed a distributing control system to integrate the three subsystems together. In a simulating under vehicle environment, the SafeBot robot successfully scanned the whole under vehicle area automatically.