Saud’s machine vision algorithm improves object recognition by using the most relevant features of an image for analysis. Photo: Victoria University of Wellington.

by Scott Kesselman

Ph.D. student Syed Saud Naqvi, from New Zealand's School of Engineering and Computer Science at Victoria University of Wellington, is working on an algorithm to help robots view images closer to how humans see.



“Right now computer programs see things as very flat — they find it very difficult to distinguish one object from another,” says Saud. 



Object detection is a complex programming problem, as many different detection algorithms exist. They can focus on patterns, color, textures or outline. Saud’s algorithm uses relevant information to decide on the type of algorithm to use in each case.



“The defining feature of an object is not always the same — sometimes it’s the shape that defines it, sometimes it’s the textures or colors. A picture of a field of flowers, for example, could need a different algorithm that an image of a cardboard box,” says Saud.



Through a scholarship, Saud will now test his algorithm on an actual robot in a real-world setting in object detection tasks in the hope that the robot will be able to successfully navigate its environment and separate objects from their surroundings.



“Most of the robots that have been dreamed up in pop culture would need this kind of technology to work,” says Dr. Will Browne, Saud’s faculty supervisor. Currently, there aren’t many home helper robots which can load a washing machine. This technology would help them do it.”



Early applications of the technology would be for tasks such as self-captioning photos on social media sites using location and other contextual clues from the image. In the future, Browne says the technology may have medical applications including being able to identify cancer cells in a mammogram.

<< Back to the News