Abstract: The operation of construction vehicles in construction and evacuation sites presents unique challenges due to the different driving conditions and surrounding environment compared to traditional transportation vehicles. Implementing autonomous driving for construction vehicles requires addressing these challenges, even though the learning approach is similar to that of cars. This thesis aims to identify suitable and highly efficient Convolutional Neural Network (CNN) models for real-time object recognition and tracking of construction vehicles, evaluate their classification performance, compare the results, and present the findings.
To achieve these objectives, a literature review and experiments were conducted. The literature review identified suitable object detection models for real-time object recognition and tracking, while experiments were performed to evaluate the performance of the selected models. Based on the literature review, Faster R-CNN model, YOLOv3, and Tiny-YOLOv3 were identified as the most suitable and efficient algorithms for detecting and tracking scaled construction vehicles in real-time. The classification performance of these algorithms was calculated and compared with each other, and the results were presented. The evaluation results indicate that YOLOv3 achieved the highest F1 score and accuracy among the algorithms, followed by Faster R-CNN. Therefore, it is concluded that YOLOv3 is the best algorithm for real-time detection and tracking of scaled construction vehicles. These findings align with the classification performance comparison reported in the literature.
Keywords: Object detection and recognition, Deep Learning, Classification performance
|
DOI:
10.17148/IJARCCE.2023.125270
