A cognitive interpretation of thermograhic images using novel fuzzy learning semantic memories

In March 2003, a severe respiratory disease have broke out in Singapore causing a total of 238 infected and 33 died, implying a fatality rate of 14%. One of the measures put in place to prevent the spread of the disease in Singapore is the employment of thermograph cameras at access points of Singapore to screen people going in and out of Singapore for fever conditions. This report looks into the technology of thermography from its basis to how it is widely used in medical fields to aid in early diagnosis of diseases like breast cancer. The objective of this final year project is to look into the possibility of creating an automated fever detection system that can aid medical staffs in times of emergency where false-negative detection of fever cases can result in very dire consequences. As we are given a set of data in the form of thermographs, image pre-processing has to be done on the image to extract the features of interest. Image segmentation methods such as Histogram-Based, Edge Detection, Template matching and Region Growing and feature extraction methods such as Gabor Feature and Scale Invariant Feature Transform will be discussed. A model of Template matching followed by Gabor Feature is proposed and implemented for the pre-processing of the images. To automate the process, we look into incorporating neural networks into the system as neural networks functions in a way that resembles the human brain. 3 neural networks were discussed in detail, namely Multilayer Perceptron, Radial basis function networks and Self-Organizing map. The proposed model is to use Self-Organizing Map to organized the input into clusters and uses it as an unsupervised tool to achieve our objective.Bachelor of Engineering (Computer Engineering