.png){kind=icon}
.png){kind=icon}
Return
Road pothole detection from smartphone sensor data using improved LSTM
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-023-16177-0.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Road abnormalities can be caused by man-made and natural disasters that affect the safety of drivers and damage vehicles. Therefore, several automatic road monitoring approaches have been proposed to monitor the road surface and detect road abnormalities like potholes. However, low accuracy in detecting the pothole in low-light conditions is taken as the main problem in this work. To address this issue, we presented an Improved Long Short Term Memory model (ILSTM) that combines a three-layer deep LSTM with a One-Dimensional Local Binary Pattern (1D-LBP) layer to detect the presence of potholes during low-light conditions and extract features such as the number of neighbouring samples and pixel values. The proposed strategy collects the pothole data from accelerometer and gyroscope sensor data using a smart phone. This model is used to classify the sensor data and label it either as a pothole or normal. Besides, this makes classification possible and extracts the location of the pothole. Evaluation results demonstrate that the proposed ILSTM approach is also robust to low lighting conditions with a detection accuracy of 99% and requires less execution time in classifying potholes and non-pothole regions on the pothole dataset collected with the help of an accelerometer and a gyroscope.
Keywords:
Road abnormalities
Long short term memory (LSTM)
One dimensional-local binary pattern (1D-LDP)
Accelerometer
Gyroscope
Pothole detection
Journal
IF:
3
Papers:
1.9W
Citations:
3.2W