Prof Maylor Leong has a vacancy for research assistant (RA) for 12 months to work in the area of computer vision and image processing.
Please contact Prof Maylor Leong at firstname.lastname@example.org for further details.
Topic: Object recognition
employing shape based indexing
The aims of this proposal are two-folds: to research on an object finder for the visually impaired (VI) and to guide the VI to reach such an object, making use of the Computer Vision technique.
For a few decades, computer scientists and engineers have attached cameras to a computer and attempt to teach a machine to see. Great success has been achieved in controlled environment when the lighting and background can be controlled. The problem remains unsolved in uncontrolled places, in particular, when objects are placed in arbitrary poses in cluttered and occluded environment. In general, object recognition is a difficult process since it needs to recognize object category. For example, a car can come with different makes and shapes. To make the matter worse, a car can also come with particular painting or pattern on its body. The recognition of a particular car is not very interesting while the recognition of the car category is more meaningful and can find more useful applications. Researchers have employed learning technique to train a classifier to tackle this category recognition problem in controlled environment. However, few have tackled this problem in cluttered environment. So far, no effective solution has yet been found.
According to our aims, we propose to research on an object finder trained to recognize at least 20 common types of objects found in cluttered environment. The proposed guidance for the VI to reach an object should investigate the matching process from object physical space to the mental space of the VI. An intelligent HCI (human computer interface) employing multi-media processing is needed. For the recognition part, there are many good approaches for single object recognition in controlled environment but there is no working technique for multiple objects recognition in cluttered environment. Here, the proposed approach is based on the idea that human can identify and recognize an occluded object from a cluttered environment, provided that enough locally perceived clues and/or distinguishable portion of the object can be observed. This observation forms the core idea for this work to test and research on. The proposed work is relevant to world emphasis on assistive technology for the elderly and less fortunate. Furthermore the researched techniques can be readily employed on a robot to support work on service robot for elder care and domestic task. This is relevant and significant to industry in Malaysia.