LOLA

Visual
Object
Detection
and
Retrieval


Background

The LOLA project is aimed at the integration of visual object recognition, orientation and retrieval components needed for achieving a degree of autonomy in mobile systems. The objective of the project is to provide a highly flexible hardware platform for the realization of this integration in an efficient and robust manner. The research activity includes issues in multivisual integration and fusion, navigation, and real-time computer architectures. The project activities consist of two parts: experimenta l verification of theoretical research issues and hardware platform development.

Sensory Processing

Artificial vision is the basis of many sensory processing techniques. In this project, a color and shape based object detection and recognition technique was developed for indoor color/shape recognition using images generated with a camera. This technique allows for objects of a desired color to be detected using the main visual camera. Once the the color is detected and the robot moves to its location, a second camera is used to determine the shape of the object by analyzing the distortion of a projected grid emmitted by an infrared laser.

Retrieval

Once the computer has determined if the color and shape of the object in question matches required parameters, the robot is equiped with a four degree of freedom arm and gripper combination that enables the robot to pick up the object and deliver it to a predetermined location.

Navigation

LOLA is equiped with sonar sensors and a tactile bumper system for navigation. The sonars determine the range of objects in every direction. Three of the sonars are mounted at mid-frame on the front of the robot for better obstacle detection in the main d irection of travel. Contact on the bumper rails stops the robot to prevent any damage to objects that are missed by the sonars.

Hardware Platform

The hardware platform consists of a Nomad 200 robot base and turret from Nomadic Inc., an Intel 80486 processor, image frame grabber with a color camera, and a laser grid emitter and polarized camera. A four degree of freedom gripper is used for object re trieval. Sixteen permanent sonar sensors are used for distance measurements.


Last modified on Apr 24, 1998, by Russell Hunt rehunt@eos.ncsu.edu