Uneven Road Surface Detection and Depth Estimation with Kinect Sensor

Lion, Kwang Ming (2018) Uneven Road Surface Detection and Depth Estimation with Kinect Sensor. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

Text Lion Kwang Ming.pdf Restricted to Registered users only Download (8MB)

Abstract

This project investigates on the uneven road surface detection and depth estimation using Microsoft Kinect sensor. The construction of speed bumps and aging roads with poor maintenance which result in large number of potholes have contributed to the uneven road surface condition. The presence of speed bumps and potholes in the middle of road have caused the car body to lose its stability when passing over it which yields a bad driving experience. Prior identification of speed bumps and potholes allow drivers and the car suspension system to respond before passing over it which may prevent bad driving experience, vehicle damages and the worst case, accident. The proposed system shows a cost effective method of speed bumps detection and height estimation using Kinect Sensor. 3D scene reconstruction was performed based on the color image and depth image acquired from Kinect’s RGB and IR cameras. Also, from the 3D point cloud, the properties of speed bump such as its distance from camera and its height were estimated. The proposed system provided reliable and consistent measurement on speed bump’s properties at accuracy of 98.93% and 86.84% for distance and height estimation respectively.