
Name: Chanitnart Mahakanta
ID: –
Title: Investigation of the Appropriate Approach to Generate the Noise Map From Crowdsourcing
Type: Thesis
Abstract
Noise
pollution is a problem that should be concerned because it leads to many
physical and mental health problems. The latest noise map in Bangkok was
released in 2001 by Department of Environment Quality Promotion. Since the
measuring processes require expensive equipment and complex procedures, it is
difficult for publishing the up-to-date noise pollution information to society.
In this study, the crowdsourcing system using smart phone to collect sound
level in the environment was proposed and also the most appropriate
interpolation method for generating sound map was investigated. To generate the
map, there are two types which are sound map and noise map. Sound map are
considered by the average of sound level in each grid for a period time. While
noise map are generated from considering the disturbance of sound level, which
exceed ten decibel A from the background of sound level in each period time. In
this study, Thammasat University, Rangsit campus is selected to be the area of
experiment for collecting data. After calibrating the mobile devices by
comparing with sound level meter with Pollution control Department office, it
found that crowdsourcing system can be applied for noise pollution monitoring.
From analyzing, the data that acquire from each devices can indicate in the
same trend even the measuring value are different. To visualize the observation
points as the area, interpolation is considered to generate the map and grid
filling is also the method which investigated. The results show that spline
method is not suitable in this study, because it calculates negative value.
Grid filling is the method that using averaging the value of observation points
in each grid. In case of no data in which grids, the grid was filled by the
value of the upper level grid size and the result maps were shown as 25 meters
grid size. For IDW, kriging and grid filling, they were evaluated by
considering RMSE. Grid filling was selected because it estimated more accuracy
of unknown value which 6.97, 6.14 and 5.14 respectively. Thus, grid filling
method was the most appropriate for generating sound map in this study.