intelligent road surface quality evaluation using rough mereology
TRANSCRIPT
INTELLIGENT ROAD SURFACE QUALITY
EVALUATION USING
ROUGH MEREOLOGY
Mohamed Mostafa Fouad
Assistant Professor at Arab Academy for Science, Technology, and Maritime TransportMember of SRGE Research Group.
Postdoctoral Fellow at VSB-Technical University of Ostrava, Ostrava, Czech Republic
Agenda
Introduction and problem domain
Aim of research
System Architecture
Data Acquisition Phase
Pre-processing Phase
Rough Mereologoy Phase
Experimental Results
Discussion
Research Obstacles
Future Work
Introduction
Smart applications nowadays are utilized to address
many common-day problems to find a convenient
solution affordable by the common citizen.
Road surface condition is an important matter in many
countries that suffer from bad road conditions.
Existence of potholes and road bumps with bad
design (homemade) can cause accidents and vehicle
damage over time
Aim of research
Provide an easy way to offer a smart distributed analysis of the road
by using a mobile application, that sends alarm for road users
before hitting road bumps or pot holes
The application measures the changes in the gravity orientation
through a gyroscope and the shifts in the accelerometer’s
indications, both as an assessment for the existence of speed
bumps.
Give the government an easy way to prioritize the process of fixing
the road conditions
Decrease the rate of accidents for vehicles’ drivers
System architecture
The architecture description:
The data acquisition phase starts by a mobile application
attached to a vehicle to detect the presence of road anomalies.
Data acquisition phase collects triple sensors values; the
Accelerometer, the Gyroscope, and the GPS. The main
intention of using the gyroscope, which represents variation
around gravity rotation, is to confirm the acceleration readings
for indicating road anomalies.
Rough mereology phase is used to rank the collected data in
order to make a useful recommendation to road user.
Pre-processing Phase
The data was collected as tuple form < Sensor Type, X-coordinate, Y-coordinate, Z-coordinate; time(in millisecond) >
The gyroscope readings have been converted from radians form into
degrees form in order to enhance the scatter point curve.
Gyroscope gravity readings around X-axis
Rough Mereology Phase
The role of the rough mereology
phase is to rank the modified data in
order to make a useful
recommendation.
The returned result of this phase is
a similarity matrix of items.
EXPERIMENTAL RESULTS
Evaluation criteria is mainly based on the computation of both the recall
and precision statistical equations.
Also with a statistical precision measurement method adopts the MAE
(Mean Absolute Error) in order to measure the recommendation quality.
EXPERIMENTAL RESULTS
Through the manual annotation of bump places we recognized that it is
always lies in the gyroscope readings in [10, 30] degrees.
According to classification rates obtained in Table
1, the precision of rough mereology in speed
bumps classification reached to 0.754, while the
recall statistical evaluation reached 0.165.
Therefore the rough mereology as a classification
algorithm provides total accuracy equals 75% with
MAE= 8.36%.
Discussion
Classification based on rough mereology does not provide promising
results as the work of [Astarita Vittorio et al. ”Automated Sensing System for
Monitoring of Road Surface Quality by Mobile Devices”], but on the other hand, it
still produces better results than the one obtained in [Mikko Perttunen et al.
, ”Distributed road surface condition monitoring using mobile phones”].
Research Obstacles Sensors readings differ from different mobile vendors (Nokia Lumia,
Nexus tablets, Samsung mobiles, …)
High error detection rate as many roads are already in a bad
structure.
Alerts of a coming speed bump will based on vehicle speed.
Readings will differ upon various vehicle models according to their
suspensions systems.
Mobile based application will suffer from high rate power
Consumption (sensing, and data transmission)
Privacy
Future Work
Try other machine learning algorithms in speed bump’s
detection process.
Support different mobile platforms.