A Sonar-based Sensor for Object Recognition

M.İ. Ecemiş and P. Gaudiano


Object recognition, sonar, robots, ARTMAP


Vision is the predominant sensory modality used for object recognition. However, machine vision algorithms are computationally expensive and require careful control of lighting conditions. In nature, vision is also dominant, but it is not the only sensory modality used for object recognition: dolphins and bats are able to perform impressive tasks using ultrasound. Sonar (ultrasonic) sensors are commonly found on mobile robots, but they are only used as a range sensor. We have developed a system for extracting information from sonar echoes to recognize objects. The system uses off-the-shelf hardware and is able to perform remarkably well under a variety of conditions. Our results show that sonar can complement vision as a sensor for object recognition.

Important Links:

Go Back