Please use this identifier to cite or link to this item:
|Title:||Color recognition through mobile applications and machine learning||Authors:||Le, Ngoc Canh||Keywords:||Engineering::Electrical and electronic engineering||Issue Date:||2022||Publisher:||Nanyang Technological University||Source:||Le, N. C. (2022). Color recognition through mobile applications and machine learning. Final Year Project (FYP), Nanyang Technological University, Singapore. https://hdl.handle.net/10356/163418||Project:||A2422-212||Abstract:||Numerous applications, such as earth remote sensing, military applications, and healthcare solutions, have benefited greatly from hyperspectral imaging . Particularly for laser emission-based detection and imaging, a better spectrum resolution has been requested. Over the years, several systems and combinations have been created to serve the ever-increasing demand. Modern cameras can attain great spectrum resolution, but they rely significantly on optical filters or a large spectrometer system to do it. As a result, a significant obstacle still exists in the capacity to extract wavelength information from typical pictures. Therefore, a robust and accurate algorithm and prediction model are developed to classify emission wavelengths based on input images. This model could potentially optimize the current process of determining wavelength of emission obtained in image and make the process become seamless and convenient. A total of 6890 images from 53 different wavelength classes captured by iPhone are used as train and test set in algorithm design and model development. Three channels of color image including red, green, and blue are explored and utilized as insightful information for algorithm design and prediction model. Machine models such as Logistic Regression, K-nearest Neighbors (kNNs), Decision Tree, Support Vector Machines (SVMs), and Gaussian Naïve Bayes are deployed to design and construct prediction model. Performance of all models are analyzed and evaluated carefully based on accuracy score, model complexity, and execution time. Eventually, the model can predict the wavelength of input images with high accuracy. Therefore, the proposed model could be deployed to use with any hyperspectral images taken with wavelength from 410 nm to 670nm.||URI:||https://hdl.handle.net/10356/163418||Fulltext Permission:||restricted||Fulltext Availability:||With Fulltext|
|Appears in Collections:||EEE Student Reports (FYP/IA/PA/PI)|
Updated on Jan 29, 2023
Updated on Jan 29, 2023
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.