A comparison of pixel-based and object-based classification methods for hyperspectral tree species classification / Muhammad Farish Danial Mohd Sham

Forests in Malaysia play a vital role in sustaining terrestrial ecosystems, making accurate tree species classification imperative for biodiversity assessment and environmental conservation. However, the challenge lies in the complexity of classifying numerous tree species within dense and diverse ecosystems. In this study, the aim is to classify forest trees using pixel-based classification and object-based image analysis (OBIA), which is supported by Hyperspectral Analysis while its objectives are to determine which classification methods are the best to use in Malaysia tropical forest, to acquire the optimum parameter for each pixel-based classification methods, assess the accuracy assessment of each classification methods, emphasizing the composition and distribution of different species. The data will be take using MicroCASI-1920 Hyperspectral VNIR Imager with Spectral Resolution Full Width Half Maximum (FWHM) less than 5nm with 5.86-pixel size and focal length of 2.5. The methodology encompasses data preprocessing, object-based image analysis, and hyperspectral signature extraction. The expected outcomes include accurate tree species maps and an overall accuracy by using and comparing several classifiers. This research's most significant consequence lies in its potential to improve forest management, biodiversity monitoring, and conservation efforts in a short time through the application of advanced remote sensing technologies, unmanned aerial vehicle (UAV) innovation and transparent methodology.