Date of Award
December 2021
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Electrical Engineering
Committee Member
Hai Xiao
Committee Member
Yunyi Jia
Committee Member
Melissa C. Smith
Abstract
Road accidents are one of the major causes of death in the United states of America. Vast majority of everyday deaths in US occur by road accidents. And poorly managed roads are one of the main reasons for it. Safety is the ultimate priority. Safety in this aspect can be ensured only with proper management of roads on a regular basis. Proper maintenance of road conditions can not only ensure safety but also increase rider comfort, fuel economy and provide better driving experience. Traditional methods of surveying is very time consuming, expensive, and require lot of human efforts. Existing methods of road surface monitoring using profilometers are very expensive. And both of these do not update the road surface conditions on a regular basis. There is a requirement of a more efficient and cost effective process to augment profilometer and road recognition systems. More than one third of the US adult population uses smart phones today. Here we present a way of dealing this problem using smart phones. Deep learning and Machine learning have step foot in almost all the industry bringing about revolutionary breakthroughs. Deep learning algorithms have many times out performed the humans especially in those problems where a pattern recognition is a key to solving the problem. Smart phones come with sensors like gyroscope, accelerometer, magnetometer, camera etc. Information from these sensors can be harnessed to detect the road conditions. This work uses this image and sensor data to detect road conditions using Deep learning algorithms. We have investigated Deep learning models on the smart phone to do the road surface detection. And in order to give heuristic and accurate information about the road surface conditions, we use cloud based collaborative approach to fuse all the data to finally update a map with these road surface conditions. Deep Learning models have been able to perform well with accuracy of 94% on sensor data model and 87.5% on vision based model.
Recommended Citation
Ramesh, Akshatha, "Cloud Based Collaborative Road Surface Monitoring Using Deep Learning and Smartphones" (2021). All Theses. 3637.
https://tigerprints.clemson.edu/all_theses/3637