JavaScript is disabled for your browser. Some features of this site may not work without it.
| dc.contributor.author | ELIYAS, ZEMENE | |
| dc.date.accessioned | 2022-03-24T06:59:12Z | |
| dc.date.available | 2022-03-24T06:59:12Z | |
| dc.date.issued | 2021-11 | |
| dc.identifier.uri | http://ir.bdu.edu.et/handle/123456789/13251 | |
| dc.description.abstract | A traffic accident is one of the leading causes of death in the world, especially in developing countries. Unsafe road conditions or road defects are one of the common reasons for traffic accidents. Detection of road defects has a big role in minimizing traffic accidents, facilitating early maintenance, and makes easier for inspections of road infrastructures. Different researches have been done on the detection of road defects and most of the studies focused on detection of the presence of damage but in the real-world scenario, the road managers need to know the type of defect, the behavior of the defect in advance in order to take actions, to allocate resources or to solve the problems earlier. In this study, a road defect detection and classification model was developed using a convolutional neural network (CNN) by considering the main types of road defects: crack, pothole, and rutting of roads. The datasets are collected in different road conditions in international repositories by cameras installed on vehicles. We make enhancement of image datasets by preprocessing images that include resizing, reduction of noises using bilateral filter and contrast enhancement with CLAHE has been made in order to enhance the image, and prepare them for the model. For feature extraction and classification, we design a CNN model with 4 way softmax classifier in order to classify the road defects to crack, pothole, rutting, and normal road conditions. The Python programming language was used to implement the model on top of the Tensorflow and Keras API, by testing and selecting the better optimizer and learning rate for the model. The model is tested using the testing dataset and its result is presented using the confusion matrix, accuracy, precision, recall and f1 score. The model achieved accuracy of 99.2% for training and 96 % for testing to detect and classification of road defects. Our model was faster to train and had smaller model size as compared to the state-of-the-art models. When the model compares with the state of the art models, it improves the performance by 2% than (VGG ) and 4% than ResNet models. Keyword - CNN, Road defect detection, defect classification | en_US |
| dc.language.iso | en_US | en_US |
| dc.subject | Communication System Engineering | en_US |
| dc.title | DEVELOPING ROAD DEFECT DETECTION AND CLASSIFICATION MODEL USING CONVOLUTIONAL NEURAL NETWORK | en_US |
| dc.type | Thesis | en_US |
