Description

Problem:

Anomalies and weather conditions on the road surface do not only affect the quality of driving, but also affect the safety of the driver, the mechanical structure of the vehicles, traffic noise, and the fuel economy. In particular, unmanned vehicles are becoming more common, and the vehicle must make decisions according to the road condition. For prediction of the road condition, several approaches have been proposed to detect problems. Some of these approaches use sensors to detect road and weather conditions. However, sensors are affected by ambient noise and therefore filtering, and signal processing techniques must be used. Apart from this, it is exceedingly difficult to detect some problems with sensors. Especially like the weather. According to the U.S. Department of Transportation's Federal Highway Administration, about 22 % of annual traffic accidents, with about 16 % of fatalities, are related to weather. In addition, most accidents were reported to have occurred in wet conditions, with 73% occurring on wet roads and 17% occurring on snow or sleet. Here, computer vision comes in. Several studies exist on road condition detection using computer vision. At some points, this study will contribute to them.

Goal:

To apply and evaluate Computer Vision methods to monitor and predict tire-road noise and road condition. The aim of the study is the monitoring and prediction of tire-road noise and weather conditions with the help of different computer vision methods. The last step will be to evaluate the performance of the applied object detection and semantic segmentation models to find the best algorithm and method.

Required Skills and Knowledge:

• Programming skills (preferably in Phyton or C++)
• Basic knowledge of data analysis, image processing and computer vision libraries (PyTorch, OpenCv, Cuda, Pandas, NumPy, ScikitLearn, TensorFlow, Keras).
• To have basic knowledge about object detection and semantic segmentation. Or the willingness to learn these things. Also, basic knowledge of image and video processing and patient annotation of images.