Road Condition Analysis for Accident Assessment Based on Unmanned Aerial Vehicle (UAV) Photogrammetry Using Image Classification

EasyChair Preprint 7240

Abstract

In recent years, many people who drove motorcycles died of traffic accidents in the world every year. As for Thailand, here also are the same situation. This paper develops an AI model of the road condition to protect motorcycle drivers by using UAV and image classification. Collect road condition images based on UAV equipment in the city area in Thailand and build the AI models by using Orthomosaic images and digital surface models based on image recognition algorithms. In the algorithm, determining the degree of risk of road conditions is one of the important steps, which used the supervised method by experts for image judgment. Image classification is used to identify the degree of road condition risk, mainly AlexNet, ResNet18 and etc. After completing the model establishment, this research will conduct a test to determine which model is more suitable for analysis by comparing the image predictions and safety factors of the two models. As a result, the RGB model is more suitable for analyzing road conditions than the DSM model and the AlexNet is more suitable for analyzing the DSM model, and the squeezenet1_0 is more suitable for analyzing the RGB model. This study aims to develop an image classification model based on road conditions in Thailand and create a mobile application to give the road condition information based on the GPS location. The model is suitable for utilizing for creating motorbike drivers’ safety guidance from UAV data.

Keyphrases: AI road, Aerial data, UAV data process, image processing, road classification

@booklet{EasyChair:7240,
  author    = {Wei Sun and Phudinan Singkhamfu and Parinya Suwansrikham},
  title     = {Road Condition Analysis for Accident Assessment Based on Unmanned Aerial Vehicle (UAV) Photogrammetry Using Image Classification},
  howpublished = {EasyChair Preprint 7240},
  year      = {EasyChair, 2021}}