Object Detection in Adverse Weather Conditions

Student Name

Hunain Hussain KhanFollow

Degree

Master of Science in Data Science

Department

Department of Computer Science

Faculty/ School

School of Mathematics and Computer Science (SMCS)

Date of Submission

Fall 2023

Supervisor

Dr. Muhammad Sarim, Visiting Faculty, Department of Computer Science, School of Mathematics and Computer Science (SMCS)

Keywords

Haze removal, Image enhancement, Object detection, Image processing, Image clarity, Computer vision, Visual perception, Image visibility, Dehazing techniques, Improved detection, Vision enhancement, Image-based detection, Atmospheric correction, Image quality improvement, Visual clarity, Haze-free images, Computer vision algorithms, Feature extraction, Image analysis and Image pre-processing.

Abstract

According to a survey carried out in Florida, between 2007 and 2016 there were almost around 1.2 million car accidents due to poor weather conditions. This resulted in around 5300 casualties during this time period. With climate change on the rise, and the increase in poor weather conditions all around the world it is necessary to have a robust system that makes commuting easier and safer for the passengers. But this effort is not just limited to systems where a human driver is involved, in fact having systems which robustly detect intrusions and objects in difficult weather conditions has a huge application in autonomous driving as well.

Heavy fog in central Pakistan results in high accident cases. A recent occurrence on Multan Road in Patuki resulted in a terrible collision involving a vehicle carrying 14 people from the same family due to risky driving conditions caused by heavy fog. This catastrophe claimed the lives of three people, including a woman, and injured 12 others.

During the recent floods in Pakistan and excessive rain, driving on long routes has become increasingly dangerous. Pakistan is the country with the 6th largest population and with this increase in population comes more vehicles on the road. According to a report carried out in 2007, every year around 7000 to 10000 people die in Pakistan every year due to road accidents. These numbers are quite alarming when viewed over a 10 or 15 year time span.

This does not only call for a system that makes commute safer during adverse weather conditions but also in general. It calls for a robust driving system which efficiently reduces the chances for accidents and increases road safety on short and long routes as well. This is where our solution comes in. We aim to develop a system which ensures road safety by efficiently and accurately detecting pedestrians, trees or other vehicles in crowded and adverse conditions such as heavy rainfall or fog.

The primary aim of this project is to develop a robust system which utilizes computer vision and deep learning to create a system which effectively detects any intrusions in difficult weather conditions and also based on the distance from the obstacle alerts the driver from a voice based system. Following list provides the description of possible intrusions that we expect our system to detect:

• Pedestrians: This may include anyone crossing the road or walking near to the premises where high vehicular density is present
• Road Signs: For an efficient object detection system it is necessary to detect road signs and adjust the speed of the car according to it.
• Trees or other objects: In order to detect trees or objects that might not be visible in windy or foggy weather, it is important for the system to detect these.
• Cars or trucks: Other vehicles that might be in the close vicinity of the car must also be detectable by the system

Document Type

Restricted Access

Submission Type

Research Project

Recommended Citation

Khan, H. (2023). Object Detection in Adverse Weather Conditions (Unpublished graduate research project). Institute of Business Administration, Pakistan. Retrieved from https://ir.iba.edu.pk/research-projects-msds/26