CNN TABANLI ALGORİTMALARIN KLASİK GÖRÜNTÜ İŞLEME YÖNTEMLERİ İLE KIYASLANARAK BİNA ÇIKARIMI PERFORMANSININ DEĞERLENDİRİLMESİ

Abstract

In the early morning of Local time at 04:17 AM, on the sixth of February 2023, (01:17 UTC), a powerful earthquake measuring Mw 7.8 struck Syria's northern and western regions as well as the southern and central portions of Turkey.. The serious problem of closed roads laden with debris, arising after the devastating earthquake in Turkey, significantly hampers disaster response and recovery efforts. This thesis emphasizes the growth and evaluation of algorithms for the detection and classification of closed roads due to debris generated after the earthquake, using images captured by drones. The study employs classic image processing methods implemented in Python along with Convolutional neural network (CNN) based algorithms. The unique dataset containing images captured after the earthquake provides a detailed picture of extensive road blockages resulting from the disaster's aftermath. An important contribution of this research the emergence of an intuitive and A GUI (Graphical User Interface) that is intuitive and easy for users to navigate. This GUI panel serves as a comprehensive platform showcasing a wide range of classical and deep learning image processing methods, allowing users to interactively visualize the results of these methods. This interactive tool not only simplifies access to algorithmic outputs but also empowers disaster response teams and decision-makers to make data-driven and effective decisions. The principal aim of of This research is to to assess how accurate and effective the proposed algorithms are in identifying the nature and prevalence of road blockages. By distinguishing between road segments filled with debris and those left open, this study provides critical information to expedite disaster relief logistics and prioritize recovery operations. The findings presented in this thesis contribute to enhancing the disaster management toolkit by expediting data-driven decision processes for post-earthquake recovery operations, thereby contributing to the overall resilience of affected communities.