Shape-based recognition using combined Jaccard and Mahalanobis measurement / Dr. Siti Salwa Salleh … [et al.]
A shape-based classifier can be built to recognize shapes using a distance measure approach. Therefore, this research is going to produce a new classifier by integrating the Mathematical equations of Jaccard and Mahalanobis. This new classifier will recognize 2-dimensional objects faster and more ac...
| Main Author: | |
|---|---|
| Format: | Research Reports |
| Language: | English |
| Published: |
Research Management Institute (RMI)
2012
|
| Online Access: | http://ir.uitm.edu.my/id/eprint/17500/ http://ir.uitm.edu.my/id/eprint/17500/2/LP_SITI%20SALWA%20SALLEH%20RMI%2012_5.pdf |
| Summary: | A shape-based classifier can be built to recognize shapes using a distance measure approach. Therefore, this research is going to produce a new classifier by integrating the Mathematical equations of Jaccard and Mahalanobis. This new classifier will recognize 2-dimensional objects faster and more accurately, which is within seconds instead of minutes. The present Mahalanobis distance measure suffers from delayed processes as its computation time formula reaches 0(n2) for n-dimensional feature's vector. This means the larger the data the longer it takes. Thus by combining the Mahalanobis math equation with Jaccard's, the expected resulted classifier will not be affected by data size. Furthermore, human drawing and sketches cannot be in similar size, similar pen pressure, clarity and orientation at all time, thus the combination of the 2 Mathematical equations above will address these issues. Secondly, the combination of 2 Mathematical equations is expected to enable the classifier detect sketches on real time with guiding assistance, to see if one's drawing is correct, as such, immediately guiding a person to re-draw. The methodology will include an analysis of Jaccard and Mahalanobis equations, studies its measurement characteristics, strengh and drawbacks. This will be followed by combining the math equations, and implementation. At the implementation stage, a data collection and analysis will be done. Data analysis consists of data filtering, feature extractions, shape normalization and segmentation tasks. Finally, the new classifier will be tested to recognize the the shapes collected earlier. The expected outcomes are i) a combined Jaccard and Mahalanobis distance measures, ii) a shape recognition classifier that can recognize shapes in seconds iv) real time checker that guide help users to draw. Continuity prospect: The classifier can be used for teaching purposes at various levels; for example for students in engineering, medicine, math and computer science mechanical drawing lesson; and also for medical therapy such as psychomotor dyslexia therapy and exercises. |
|---|